From Artificial Intelligence to Aerospace Medicine, These Rising Stars Are Shaping the Future

The Technion is proud to celebrate the inclusion of four exceptional students and alumni on this year’s Forbes Israel 30 Under 30 list. Their groundbreaking achievements span artificial intelligence, space medicine, and deep-tech innovation—each one a shining example of how Technion graduates are making a global impact.

Dr. Dean Leitersdorf CEO & Co-founder, Decart.ai | Age: 26

Dean Leitersdorf isn’t just dreaming big—he’s building big. As co-founder of Decart.ai, Dean is on a mission to create a trillion-dollar AI company that could rival tech giants like Google and TikTok. A triple Technion graduate of the Taub Faculty of Computer Science with a PhD by age 23, Dean previously served in an elite IDF Unit and won the Israel Defense Prize.

Decart’s AI efficiency platform is already disrupting the market, and its AI-powered game Oasis reached 1 million users in just three days—faster than ChatGPT. With $53M in VC funding and profitability in its first year, Dean’s bold vision is just getting started.

“If you’re not in the top 0.1%, it’s not interesting.”

Dr. Summer Sofer Founder, Israeli Society for Aerospace Medicine | Age: 29

From the soccer pitch to NASA, Dr. Summer Sofer is breaking boundaries. A black belt in karate and former player on Israel’s national soccer team, she’s now pioneering space medicine in Israel while completing her medical degree at the Technion’s Rappaport Faculty of Medicine.

Born in New York and raised on resilience, Summer founded the Israeli Society for Aerospace Medicine to grow this underdeveloped field at home. She’s currently completing a specialization at NASA and envisions a national infrastructure for space medicine in Israel.

“It’s easier to advance something you truly believe in.”

Hen Davidov Rhodes Scholar & AI Researcher | Age: 25

“I never thought I fit the profile of a Rhodes Scholar,” says Hen Davidov—yet he’s now one of only two Israelis selected this year. His Technion-based research blends AI and medicine, focusing on building trustworthy diagnostic systems that support doctors with clear probability-based predictions.

Inspired by personal experiences with family illness, Hen’s work is already helping refine breast cancer diagnostics. A graduate of the Taub Faculty of Computer Science and soon to begin his PhD at Oxford, he aims to set new global standards for ethical, reliable medical AI.

“When it comes to medicine, the risks are multiplied.”

Dr. Ameer Haj Ali Founder, Universal AI | Age: 29

Dr. Ameer Haj Ali is redefining what’s possible in deep tech. Raised in Shfaram, a graduate of the Viterbi Faculty of Electrical and Computer Engineering, Ameer completed a record-fast PhD at UC Berkeley in just two years.

In 2025, he launched Universal AI—an ambitious startup focused on the next generation of AI infrastructure. Within two months, the company secured $10M in funding from high-profile investors, including Eric Schmidt, Jared Kushner, and Elad Gil. Ameer’s tireless commitment (including sleeping in the office!) reflects a deep drive to turn bold ideas into real-world impact.

“Time is the most precious resource. I feel behind every day.”

These four honorees represent the best of Technion’s spirit: fearless ambition, technical brilliance, and a commitment to solving real-world challenges. We salute their achievements—and can’t wait to see what they build next.

A new interdisciplinary study by researchers from the Ruth and Bruce Rappaport Faculty of Medicine and the Andrew and Erna Viterbi Faculty of Electrical and Computer Engineering at the Technion reveals a surprising insight: local release of dopamine—a molecule best known for its role in the brain’s reward system—is a key factor in acquiring new motor skills

From writing and typing to playing a musical instrument or mastering a sport, learning movement-based tasks is one of the brain’s most complex challenges. This collaborative new study reveals how the brain reorganizes its neural networks during such skill learning and uncovers the vital role of dopamine in this process of motor learning.

The research, published in Nature Communications, was led by Dr. Hadas Benisty, Prof. Jackie Schiller, and M.D./Ph.D. student Amir Ghanayim, with contributions from Prof. Ronen Talmon and student Avigail Cohen-Rimon from the Andrew and Erna Viterbi Faculty of Electrical and Computer Engineering.

The ability to acquire new motor skills is fundamental for adapting to our environment. This learning takes place in the primary motor cortex—a region of the brain responsible for planning and executing voluntary movements. From this cortical “command center,” signals are sent via the spinal cord to activate muscles and coordinate movement. Neural activity in this region is known to change as we learn new skills. However, the mechanisms that drive these changes remain unclear.

Key findings of the study

The researchers used advanced calcium imaging in behaving mice and chemogenetic inhibition techniques—engineered receptors and specific drugs—to temporarily switch off targeted brain cells, allowing researchers to study their function. They mapped dynamic changes in neural networks with cellular resolution within the motor cortex during the acquisition of a motor skill, and discovered that during learning, neural networks transition from a “beginner” to an “expert” structure.

Crucially, this process depends on the local release of dopamine in the motor cortex. Under normal conditions, dopamine molecules are delivered to this region by neurons originating in the ventral tegmental area (VTA)—a central dopamine hub in the brain. The researchers hypothesized that this dopamine release triggers plasticity mechanisms, leading to changes in functional connectivity between neurons in the motor cortex. This process enables motor learning by storing new skills for future use. In essence, this is a form of reinforcement learning, where successful movement outcomes reinforce the brain’s internal wiring.

What happens when dopamine is blocked?

To test the necessity of this mechanism, the researchers examined both the activity and functional connectivity of the neural network and the learning process when dopamine release in the primary motor area was blocked. The results were clear: When dopamine was blocked, learning stopped completely—mice were unable to improve their performance in a forelimb-reaching task. The motor cortex neural network remained static. However, as soon as dopamine release was restored, learning resumed, along with reorganization of the neural network.

The study provides compelling evidence that local dopamine release serves as a crucial signal for neural plasticity in the motor cortex, enabling the necessary adaptations for producing precise and efficient motor commands. A particularly interesting discovery was that blocking dopamine did not affect previously learned motor skills. In other words, the researchers proved that dopamine is essential for learning new movements but is not required for performing already learned ones.

This study represents another step toward understanding brain plasticity and learning mechanisms at the cellular and network levels. It highlights the brain’s ability to reorganize itself, allowing us to refine our motor skills throughout life. These insights may also have important implications for treating neurological disorders such as Parkinson’s disease, where dopamine production is impaired, and motor learning is compromised.

The laser missile defense system will be managed by the newly established Energy Warfare Administration in Rafael’s Land and Naval Systems Division.

With Israel’s Iron Beam laser missile defence system to become operational in the fourth quarter of 2025 Rafael Advanced Defense Systems today announced the establishment of a new Energy Warfare Administration in its Land and Naval Systems Division. The new administration will manage high-power laser systems projects.

The new administration head named only as Dr. Y. will serve under Rafael EVP & General Manager of the Land & Naval Systems Division Tzvi Marmor. Dr. Y., a graduate of the Technion with a Ph.D. in physics, joined Rafael about 12 years ago and has since held a series of senior positions in the engineering sector while leading the development and production of groundbreaking systems for national security. Among other things, she served as head of Israel’s Iron Dome production line and as head of electro-optics, which is integrated into most of the company’s advanced systems.

Rafael’s Land and Naval Systems Division at Rafael will continue to be responsible for the development, production and marketing of complete and integrative products and solutions in the areas of precision attack, including the Spike missile family, active, reactive and passive defense systems for tanks and armored vehicles, including Trophy, and Iron Beam.

The new administration will be responsible for Iron Beam in particular and the development and production of Rafael’s laser systems in general.

Rafael CEO Yoav Tourgeman said, “Dr. Y. brings with her extensive management experience of teams of hundreds of developers, as well as a deep understanding of technological needs and operational requirements. All of this will allow her to promote these flagship projects and realize the marketing and business potential in Israel and around the world.”

High-power laser system for ground-based air defense

Iron Beam is a high-power laser system for ground-based air defense, against aerial threats (rockets, mortar bombs, drones, and cruise missiles). The Ministry of Defense Directorate of Defense R&D (DDR&D) (MAFAT) is leading the project, along with Rafael, the main developer, and Elbit Systems.

Iron Beam be integrated into Israel’s multi-layered air defense system, alongside Iron Dome, which is an interception system for rocket threats within a range of 40 kilometers. Iron Beam will be a complementary system for intercepting rockets within a range of up to 10 kilometers, using a powerful 100 kilowatt laser beam. The major advantage of Iron Beam is in significant cost savings. While each interception with Iron Dome costs an estimated $30,000, each interception with Iron Beam will cost just $5-$10.

TEL AVIV, Israel, May 19, 2025 — Quantum Machines, the leading provider of advanced hybrid quantum-classical control solutions, announced today the release of QUAlibrate, an open-source framework for calibrating quantum computers.

The framework dramatically shortens calibration times and provides a comprehensive solution for creating, executing, and sharing calibration protocols across different quantum computing platforms. By creating an open ecosystem, QUAlibrate enables researchers and companies worldwide to build upon each other’s advances, accelerating the path to practical quantum computers.

“QUAlibrate has been transformative for our company,” said John Martinis, CTO and co-founder of Qolab. “Its automated calibration capabilities now complete full calibrations in less than 10 minutes – tasks that otherwise would demand up to two hours of manual work. This efficiency boost frees up our team to focus on accelerating our QPU development.”

Calibration has emerged as one of the most critical bottlenecks in scaling quantum computers. To properly initialize and maintain a quantum computer’s performance, calibration must be performed not just once, but frequently during operation to compensate for system drift. As quantum systems grow in size, the calibration challenge becomes exponentially more complex. For instance, calibrating a 100-qubit superconducting quantum computer from scratch can take up to two days, and even recalibrating an already-calibrated system can take an hour or more. This becomes impractical when scaling to future systems with hundreds of thousands of qubits.

“We care both about how long it takes to calibrate and about how good the calibration is, two things that sometimes collide, and this impacts the performance of the quantum computer as a whole,” said Dr. Yonatan Cohen, co-founder and CTO of Quantum Machines. “We built an open-source solution because we believe this is a challenge the community can solve together. Researchers in both academia and industry continuously develop new calibration algorithms and protocols. One day, a team in Boston might develop a protocol that increases quantum operation fidelity, the next day a European company might create a method to speed up calibrations. The path to solving this fundamental challenge lies in a collaborative approach where teams can instantly leverage each other’s advances and build on them.”

To address this fundamental challenge, Quantum Machines has developed QUAlibrate, an open-source calibration framework that transforms quantum calibration from a collection of isolated scripts into a modular, collaborative system. QUAlibrate enables researchers and quantum engineers to create reusable calibration components, combine them into complex workflows, and execute calibrations through an intuitive interface. The platform abstracts away hardware complexities, allowing teams to focus on quantum system logic rather than low-level details.

In a recent demonstration at the Israeli Quantum Computing Center (IQCC), QUAlibrate completed a multi-qubit calibration of superconducting qubits in just 140 seconds. The result demonstrates the system’s speed and efficiency in real-world conditions.

QUAlibrate’s open-source nature and modular architecture mean that when researchers develop new calibration protocols, these innovations can be immediately shared, validated, and built upon by the broader quantum computing community. Companies can also develop proprietary solutions on top of QUAlibrate that leverage advanced approaches like quantum system simulation and deep learning algorithms. This creates an ecosystem where fundamental calibration advances can be shared openly and enables specialized tools that push the boundaries of performance.

“It was fantastic to see QUAlibrate rapidly perform a complex, full tune-up on the Architect system, OQC’s partner system with OINS and Quantum Machines,” said Simon Philips, CTO of Oxford Quantum Circuits (OQC). “The results clearly demonstrated the power and efficiency of QUAlibrate’s automated calibration approach, as showcased by Quantum Machines.”

“At Quantum Elements, we see QUAlibrate as a meaningful step toward a more open and empowered quantum ecosystem,” said Izhar Medalsy, co-founder and CEO of Quantum Elements. “Calibration has long been a hidden bottleneck, often locked behind proprietary tools and inaccessible workflows. By making it open source, Quantum Machines is helping turn calibration into a shared foundation the entire field can build on. We believe this kind of openness not only accelerates progress — it also gives scientists the clarity and control they need to push quantum computing forward, together.”

“QUAlibrate has laid a vital groundwork for fast, reliable and efficient calibration on our QPU while we continue to scale up the size, connectivity and fidelity,” said David T. Lee, Research Scientist at Academia Sinica. “It’s definitely a game changer.”

Along with the framework, Quantum Machines is releasing its first calibration graph for superconducting quantum computers, providing a complete calibration solution that can be immediately deployed and customized. The graph leverages QUAlibrate’s parallel calibration capabilities to dramatically reduce calibration times. Looking ahead, Quantum Machines and NVIDIA are developing software libraries that will integrate QUAlibrate with accelerators like the NVIDIA DGX Quantum, enabling even faster calibration times and higher fidelity calibrations using machine learning models.

Quantum computing researchers and engineers can begin using QUAlibrate today by accessing the open source repository at: https://github.com/qua-platform/qualibrate or visit Quantum Machines’ website to learn more: https://www.quantum-machines.co/products/QUAlibrate.

About Quantum Machines

Quantum Machines (QM) is a leading provider of quantum control solutions, driving the advancement of quantum computing with its Hybrid Control approach. By harmonizing quantum and classical operations, Hybrid Control eliminates friction and optimizes performance across hardware and software, enabling researchers and builders to iterate at speed, resolve setbacks, and bring visionary ideas to life. Its platform supports any type of quantum processor, empowering the industry to scale systems, accelerate breakthroughs, and push the boundaries of what’s possible.

Medical robotics first entered general surgery in the 1980s with laparoscopic tools that enabled minimally invasive procedures, reducing incision size and recovery time. These early systems extended surgeons’ capabilities, transforming the surgical landscape.

Today, artificial intelligence (AI) is ushering in a new era of precision and control in the operating room. Yet despite this progress, robotic systems remain limited to select procedures, leaving most surgeries dependent on traditional methods — and many patients without the benefits of enhanced consistency and outcomes.

As medical technology continues to evolve, how can AI applications in surgical robotics scale to transform healthcare on a broader level?

Fueled by increased robotic VC funding and the digital transformation of the last five years, the robotics industry is seeing fast-tracked market results with no signs of stopping. Earlier this year, Nvidia announced its intent to increase investments in its robot development, signaling a positive shift for the future of robotics. Similar investments in robotics by large-scale players will further advance robotic technology through data collection and machine learning, while providing additional resources and insights.

Surgical robotics industry leaders, such as Intuitive Surgical, Medtronic, and Stryker, have pioneered robotic-assisted surgeries for various procedures. Since introducing its da Vinci system for general surgery in 2000, Intuitive Surgical has continued to iterate its robotic platform to expand its offerings to cardiac, bariatric, gynecology, and thoracic surgeries, among others. With the mass adoption of robotic-assisted surgeries, surgical robotics have consistently been adopted at a faster scale. Between 2012 and 2018 alone, robotic-assisted procedures rose 738% in general surgery.

Looking ahead, surgical robotics have even greater market potential, and are predicted to grow to over $14 billion by 2026 – up from just over $10 billion in 2023. This is mainly due to greater access to robotic surgery procedures, advancements in automation and digital technologies, and new players who aim to deliver cutting-edge medical solutions that harness the power of AI.

Deep Tech Approach

Built on the intersection of disciplines, deep tech merges multidisciplinary technologies such as AI, quantum computing, biotechnology, and robotics to usher in a new era of technology. Startups embracing a deep tech approach in robotic surgery are creating innovative solutions for the future, as can be seen in healthtech development, which can improve patient access to critical medical care. With deep tech development, surgical procedures may become fully automated down the road, requiring minimal surgeon assistance and significantly expanding access to treatment.

Emerging deep tech technologies in surgical robotics can make a lasting global impact. With roughly two-thirds of the worldwide population – 5 billion people – lacking access to surgical treatment, these new modalities, powered by AI, can expand general access and close the surgical care gap.

Fusing AI and Surgical Robotics

AI has innovated and changed how we interact with different technologies and each other. Over the last five years, the transformation brought on by AI has accelerated the development of robotics and created additional applications for AI within different modalities, including robotic surgery.

Here are three essential ways AI is making a fast and profound impact:

1. Embodied AI

Technology is changing how we interact with our environment and the people around us. Embodied AI, which includes autonomous vehicles and humanoid robots, is the fusion of AI with physical systems to execute complex tasks in real-world settings. When embodied AI is applied to surgical robotics, it has the potential to have long-lasting impacts on enhancing surgical care and improving existing techniques. However, embodied AI requires significant real-world data to develop training simulation models, which are used to train and expand AI capabilities and improve data-driven insights. Until recently, access to large amounts of training data has been somewhat limited; however, as the industry continues to invest in the training and development of AI models, the simulated data pools are growing at a quicker pace and improving embodied AI functionality.

2. Continuous Data Insights and Guidance

AI-based systems can absorb and comprehend large swaths of information in seconds – much faster than the human brain. By training machines on large data sets, data-driven insights can inform surgical decisions before surgeons even set foot in the OR. AI-driven training simulations can significantly benefit surgeons, as training on data sets that are based on thousands of surgeries provide surgeons with trends and techniques to deliver a better patient experience, and also allow them to prepare for and understand the intricacies of rare or complex cases before they face them in the OR. This process can significantly accelerate and shorten the long learning curve surgeons face when training to reach peak surgical performance.

When applied to real-time imaging and visualization technologies, AI-driven data can also enhance surgeons’ decision-making capabilities during operations. By providing surgeons with insights to adjust surgical plans during procedures, AI-based systems can empower surgeons to optimize techniques and approaches in real-time. Through AI-driven imaging systems, surgeons can receive advanced imaging analytics and real-time 3D “maps” of the surgical sites. These augmented overlays can give surgeons expanded insights into the surgical field alongside real-time feedback on their surgical techniques. Robotic surgery platforms are at the forefront of integrating this technology into the OR, with the goal of increasing surgical precision and outcomes.

Furthermore, by providing ongoing feedback post-operation, AI-based systems can provide valuable feedback to surgeons about their performances during procedures – highlighting weaknesses and strengths, and suggesting specific strategies on how to improve them. Such platforms can also recommend new treatment plans based on each patient’s history and the particular procedure’s data analysis, and empower surgeons with additional information that can enhance further treatment. As such, AI platforms have the potential to absorb and adapt surgical feedback throughout the full surgical cycle (before, during, and after) through an AI feedback loop to increase surgeons’ precision and performance.

3. Increased Accuracy and Precision

Individual surgical skills often vary among surgeons due to their access to top-tier opportunities, from program location to surgical mentorship access. For instance,  the field of ophthalmology has a steep learning curve. On average, it takes at least 15 years of training and surgical experience to reach peak performance as an ophthalmic surgeon. With a growing aging population and a dwindling number of surgeons, a new solution is needed to reduce the surgeon’s training period and standardize the accuracy and precision of care for all.

In addition to reducing the learning curve for surgeons and allowing them to reach peak performance faster, introducing AI-based platforms into the surgical process can increase accuracy and precision and may improve suboptimal outcomes. Semi-autonomous and increasingly autonomous features in robotic platforms can eliminate the surgeon’s natural hand tremor and improve overall precision and accuracy, thus improving clinical outcomes. In addition, the ability of AI-based systems to recognize unique anatomical structures and provide the exact location for incisions and other surgical steps – especially in complex procedures or anatomical areas – can significantly reduce the rate of surgeon errors by improving spatial awareness of anatomical structures. As such, all surgeons utilizing AI-based systems will be able to provide consistently more precise care.

When incorporated into the surgical process, AI-based robotic platforms provide invaluable insights that can enhance the overall experience for both the patient and the surgeon.

Conclusion

AI will continue to play a significant role in advancing healthcare in the future. Incorporating advanced AI technologies into our healthcare services, such as electronic filing, diagnostics, and health monitoring and tracking, as well as surgical care, is imperative. In doing so, we can improve the overall patient and surgeon experience.

In robotic surgery, AI  is expediting the technology’s transformation and patient access to consistent, high-tier treatment. Advancements in robotics, coupled with AI and automation, will continue to usher in new applications, creating a higher level of standardised care and launching healthcare quality and access to new heights.

A new study introduces choice engineering—a powerful new way to guide decisions using math instead of guesswork. By applying carefully designed mathematical models, researchers found they could influence people’s choices more effectively than relying on gut instincts or even traditional psychology. This discovery could pave the way for smarter, more ethical tools to improve decision-making in areas like education, health, and everyday life.

The new study, published in Nature Communications, demonstrates that mathematical models can be more effective than psychological intuition when it comes to influencing human decisions. Led by Prof. Yonatan Loewenstein from Safra Center for Brain Sciences (ELSC) at Hebrew University, in collaboration with Dr. Ohad Dan from Yale University and Dr. Ori Plonsky from the Technion, the research introduces a novel concept: choice engineering.

The study draws a distinction between two approaches to influencing behavior. The first, known as choice architecture, has gained widespread popularity since one of its pioneers, Richard Thaler, was awarded the Nobel Prize in Economics in 2017—with behavioral insights (“nudge”) teams emerging in governments around the world.

Choice architecture relies on psychological principles—such as primacy, anchoring, or intuitive heuristics—to subtly steer decisions. The second approach, proposed by the researchers, is choice engineering: a method that uses computational models and optimization techniques to systematically shape behavior with precision.

To put these approaches to the test, the team launched an academic competition where international academic teams were tasked with designing an incentivization mechanism (“reward schedule”) that would get people to choose one of two objectively equal-value options.

More than 3,000 participants took part in the experiment, each exposed to one of several reward strategies. Some were built on intuition and psychological insights, while others were crafted using computational models.

The most effective schedule was based on a computational model called CATIE (Contingent Average, Trend, Inertia, and Exploration), designed by Dr. Ori Plonsky together with Prof. Ido Erev from the Technion. The model integrates multiple behavioral tendencies into a unified predictive framework. This CATIE-based strategy significantly outperformed those based on the widely used machine-learning model Q-learning, and those informed by qualitative intuition alone.

“Our study shows that just as engineers use mathematical models to build bridges or design aircraft, we can use models of learning and decision-making to influence behavior—reliably and efficiently,” said Prof. Loewenstein.

The findings demonstrate that behavior can be engineered with surprising accuracy when guided by well-calibrated models. Moreover, the study offers a new method for evaluating cognitive models—not only by their explanatory power, but also by their effectiveness in shaping real-world decisions.

The implications are far-reaching. In fields ranging from education and public health to digital design and policy-making, choice engineering could enable the development of empirically optimized, scalable interventions. At the same time, the researchers note that ethical frameworks will be essential to guide the responsible application of these tools.

As a proof of concept, this study underscores the emerging potential of mathematical modeling in the cognitive sciences—not just for understanding behavior, but for actively guiding it.

The AI-powered inspection system will complement manual inspections to improve efficacy and boost customer experience.

Hertz, one of the world’s leading car rental companies, has partnered with Israeli AI-driven vehicle inspection systems UVeye to enhance its vehicle maintenance practices in the United States.

Vehicle inspections in the rental industry have always relied on manual inspections. AI-based maintenance processes enhance the accuracy and efficiency of the procedure.

UVeye’s AI-powered camera systems and machine learning algorithms enable real-time automated inspections of a car’s body, glass, tires and undercarriage. The system has been installed at hundreds of dealerships, fleet sites and auction lots globally.

Amir Hever, CEO and cofounder of UVeye, said its AI-powered inspection systems “complement manual checks with consistent, data-backed assessments completed in seconds.”

The partnership will also enhance Hertz’s service and vehicle availability thanks to Uveye’s proactive ability to detect maintenance issues quickly and precisely. For instance, the system captures and instantly analyzes high-resolution images of the tire treads to determine whether a tire needs to be replaced.

Hertz’s EVP Technical Operations Mike Moore said: ‘With millions of customers and over 100 years of service around the world, we’re continually focused on transforming every aspect of our company and that includes how we maintain our vehicles.”

During the initial rollout, the technology will be deployed at Hertz locations across major US airports, starting with Atlanta’s Hartsfield-Jackson International Airport, the first to be equipped with UVeye systems. Deployment is expected to be completed by the end of the year.

In 2024, UVeye was named as one of TIME’s “Best Inventions” and Fast Company’s “World’s Most Innovative Companies.” As of February, the company had raised a total of $380.5 million in capital. 

The company is developing DeltaStem, an AI-driven platform designed to improve the production of human cells for therapeutic use.

Somite AI, a biotechnology company developing AI tools for human cell therapy, has raised over $47 million in a Series A funding round led by Khosla Ventures. Other participants include Max Levchin’s SciFi Ventures, The Chan Zuckerberg Initiative, Fusion Fund, Ajinomoto, Pitango HealthTech, TechAviv, Harpoon Ventures, along with angel investor and former Chairman of Recursion, Dr. R. Martin Chavez.

The company, which once called itself “the OpenAI of stem cell biology”, has also welcomed its new CEO of Applications and Board Member Fidji Simo as an investor. Earlier investors included Texas Venture Partners, and this new round brings its total funding to roughly $60 million.

Somite AI is developing DeltaStem, an AI-driven platform designed to improve the production of human cells for therapeutic use. The new funding will advance its capabilities and support programs targeting Type 1 Diabetes, orthopedic injuries, muscular diseases, and blood disorders.

“We’re building the foundation model for the human cell,” said Founder and CEO Dr. Micha Breakstone. “By generating the world’s largest cell signaling dataset at 1000x the efficiency of current methods, we’re training DeltaStem to deliver protocols with unmatched purity, scalability, and reliability. We are rapidly driving towards an AlphaFold moment for developmental biology, enabling the scalable production of any cell, for anyone.”

Somite AI’s capsule technology generates large-scale cell state transition data that feeds into the DeltaStem model. According to the company, this allows faster development of cell differentiation protocols compared to traditional approaches.

“I think we’re really at a dawn of a new age where we’re really using or leveraging AI to usher in this new age or era of human regeneration and repair,” Breakstone added, in an interview with CTech. “I think if you believe in AI and the exciting opportunities that it yields and the ability to make us more creative, smarter, more intelligent, more prosperous, I think the next frontier is actually not resigning ourselves to letting our own body deteriorate. I think that is what Somite is about: to come in and replenish the body with our own types of cells, but any type of cell to cure diseases. That’s the next level of prosperity and abundance that we want to be ushering in.”

Somite AI was co-founded by Breakstone, a serial AI entrepreneur, and Dr. Jonathan Rosenfeld, Head of the Fundamental AI Group at MIT. Other co-founders include Harvard and University of Washington researchers Prof. Olivier Pourquié, Prof. Allon Klein, Prof. Jay Shendure, and Prof. Cliff Tabin.

“Traditional cell therapies are expensive, slow to develop, and unpredictable. AI can systematically solve these challenges,” added Vinod Khosla, founder of Khosla Ventures. “Somite AI’s foundation models, once fully developed and validated, will not only create value for their own pipeline, but have the potential to reshape the entire field of human cell therapy.”

The early-stage startups will benefit from a three-month partnership with the semiconductor giant and other multinational corporations.

Ten of Israel’s brightest new hopes in the world of deep tech – complex innovations to solve the world’s toughest problems – have been selected for an intensive nurturing program.

The early-stage startups will benefit from a three-month partnership with the semiconductor giant Intel and other multinational corporations.

They were selected by Ignite DeepTech — a new independent accelerator inspired by Intel’s Intel Ignite initiative — from 258 applicants working to drive significant change in many sectors: cybersecurity, AI applications and infrastructure, data and cloud infrastructure, biotechnology, drones and robotics, particle accelerators, quantum technologies, developer tools and road safety.

Ignite DeepTech is supported by the Israel Innovation Authority and Israel’s Economy Ministry.

The Ignite DeepTech startups benefit from the hands-on guidance of experienced entrepreneurs, who don’t take equity. They also receive tailored, intensive support focusing on product-market fit, preparation for advanced funding rounds, looking after their mental health, product development, business strategy, marketing and customer management.

Many of the 10 chosen startups are still in stealth mode – operating under the radar while they develop their technology, to avoid tipping off competitors.

But here’s what we do know about them:

  • SkyPulse Technologies – Fast, agile, flexible made-in-Israel drones with high-end capabilities, designed for affordability and versatility in critical missions.
  • DYM sense – Revolutionising road safety with noninvasive alcohol detection technology that prevents drunk driving.
  • Impala.ai – A platform that allows companies to run large AI models and process vast datasets efficiently, making high-performance AI accessible and scalable for businesses.
  • MNDL Bio – AI-powered solutions that optimize gene expression and significantly increase protein production yields for companies that use genetic engineering.
  • DataFlint – A user-friendly platform that helps organizations using Apache Spark (for big data analytics) to quickly identify and fix performance bottlenecks in their systems.
  • Particle Lab – Pioneering a new architecture for particle accelerators (think Large Hadron Collider, in Geneva, but generally much smaller).
  • Troup AI – An LLM (large language model) inference platform, which means it infers, rather than relying on being fed data to be trained.
  • Twine Security – AI-powered digital employees, including one called Alex, who can perform cybersecurity tasks instead of humans.
  • Huskeys – An AI-powered security platform that defends against sophisticated, dynamic cyberattacks.
  • Jazz –  Cybersecurity.

Alon Leibovich, managing director of Ignite DeepTech, said: “We expect the deep-tech sector to experience explosive growth in the coming years, tackling real-world challenges like spacecraft, robotics, energy, and more.

“We’re excited to support the trailblazing startups selected for this program. Alongside our new programs for pre-seed startups and deep-tech executive training, this brings us closer to realizing our vision of a full support platform for Israel’s deep tech industry.”

Nature Reviews Clean Technology spotlights Decoupled Water Electrolysis (DWE) – a novel approach to green hydrogen production pioneered by H2Pro that solves key challenges in direct connection to solar and wind.

For decades, water electrolysis has remained stagnant, relying on conventional technologies like alkaline and PEM, where ongoing development yields only incremental gains in overcoming the barriers to affordable green hydrogen production. Now, a new category is gaining global recognition: DWE – an approach that tackles these challenges with fresh thinking. At the center of its rise is Israeli climate tech company H2Pro, whose bold reimagining of electrolysis is featured in a landmark review in Nature Reviews Clean Technology.

The article highlights a critical challenge: conventional electrolyzers struggle to operate safely and efficiently under fluctuating solar and wind power. Membranes, gas crossover risks, and operational constraints limit their ability to respond dynamically to intermittent renewable energy, driving up costs and limiting deployment.

“To unlock the full value of cheap renewable electricity, we need electrolysis that can go behind the meter and be fit for green – hyper-flexible, ultra-low cost, seamless on/off, and efficient across a wide range of power loads,” said Rotem Arad, CBO of H2Pro and article contributor. “By splitting hydrogen and oxygen into two distinct steps, mediated by a proprietary redox cycle, that’s exactly what H2Pro’s DWE does.”

The review was co-authored by Prof. Avner Rothschild and Dr. Guilin Ruan (Technion – Israel Institute of Technology), Dr. Fiona Todman and Prof. Mark D. Symes (University of Glasgow), Dr. Tom Smolinka (Fraunhofer-Institut für Solare Energiesysteme ISE), Prof. Jens Oluf Jensen (DTU – Technical University of Denmark), Gilad Yogev and Rotem Arad (H2Pro). Together, they examine the chemistry, system architectures, and commercial implications of decoupling hydrogen and oxygen — and validate growing consensus that DWE could be key to scaling green hydrogen cost-effectively.

“When we conducted the groundbreaking Technion research that became the foundation for H2Pro, we knew incremental improvements to legacy electrolysis weren’t enough,” said Dr. Hen Dotan, CTO and co-founder of H2Pro. “We let go of outdated assumptions — like the belief that hydrogen and oxygen must be produced simultaneously — and ended up pioneering not just a breakthrough technology, but a new mindset around electrolysis. We’re thrilled to see DWE gaining momentum and honored to be featured alongside the esteemed researchers advancing the field.”

H2Pro is now preparing to deploy the world’s first decoupled electrolysis system in the field — a major step in translating science into scalable commercial infrastructure. Scheduled for installation this year in Tziporit, Israel, it will also be the country’s first green hydrogen project.