Suggestions
Vladislav Voroninski
Russian-American mathematician and entrepreneur
Vladislav Voroninski, CEO of Helm.ai
Vladislav Voroninski is a Russian-American mathematician, entrepreneur, and the CEO and co-founder of Helm.ai, an AI software company focusing on autonomous driving.12
Academic Background
- Received B.S. and M.A. degrees in Applied Mathematics from UCLA in 2008, summa cum laude1
- Earned Ph.D. in Mathematics from UC Berkeley in 2013 under Emmanuel Candes and John Strain1
- Was on the faculty at the MIT Mathematics Department from 2013 to 201612
Research Contributions
- His PhD thesis kicked off the study of phase retrieval in applied mathematics, providing the PhaseLift algorithm and recovery guarantees1
- Research has led to solutions in computer vision, quantum operator theory, optimization, deep learning, and compressive sensing1
- Connected deep learning and inverse problems, resolving the sample complexity bottleneck for compressive phase retrieval1
Awards and Honors
- Awarded the 2014 SIAM Outstanding Paper Prize for originality and opening up new areas of applied mathematics12
- Received the Bernhard Friedman Memorial Prize from UC Berkeley for his PhD thesis1
- Recipient of the George E.G. Sherwood Prize from UCLA Mathematics Department in 2008 for being the top graduating senior12
Entrepreneurship
- Founding Chief Scientist at Sift Security, a cybersecurity startup acquired by Netskope in 201812
- CEO and co-founder of Helm.ai since 2016, building AI software for autonomous driving12
Helm.ai is developing full-stack AI software for ADAS and L4 autonomy using unsupervised learning techniques like Deep Teaching.34 The company has raised $55 million in funding and works with global automakers and Tier 1 suppliers.3
Highlights
"Equity Portfolios Can Hedge Smart-Car Risks: Diversifying investments across critical suppliers could balance concentration risks. A portfolio that includes companies such as Applied Intuition, Helm.ai and Momenta mirrors Tesla’s core autonomous capabilities."
Nice write-up by Bloomberg Intelligence on:
The huge market opportunity high end ADAS poses in automotive, Tesla's lead amongst OEMs, the role of generative AI in accelerating progress in autonomous driving, and the potential for OEMs to challenge Tesla by partnering with certain suppliers to achieve self-driving breakthroughs.
Couldn't agree more! We will soon see automakers catching up to and even surpassing Tesla by securing critical AI partnerships.
https://lnkd.in/gF9CXpAB
I'm thrilled to announce the launch of WorldGen-1, a first of its kind multi-sensor generative AI foundation model for simulating the entire autonomous vehicle stack.
WorldGen-1 outputs highly realistic AI-generated sensor data across multiple modalities simultaneously, including multi-view camera, semantic segmentation, Lidar birds-eye-view, Lidar front-view, and ego-vehicle path.
To create WorldGen-1, we combined innovation on generative AI DNN architectures with our Deep Teaching unsupervised training technology, which together result in a highly scalable and capital efficient form of generative AI.
All the generated data is self-consistent, modeling complex relationships between sensors, both geometrically and in terms of appearance. The resulting simulation enables highly streamlined training and validation of autonomous driving systems, especially for resolving rare corner cases that are otherwise extremely costly to collect in the wild.
WorldGen-1 is also able to extrapolate solely from real camera data a high fidelity simulation of the rest of the autonomous driving stack, including multi-view Lidar and the ego vehicle path in physical coordinates. This capability demonstrates just how information-rich the camera sensor is and the power of generative AI for sensor simulation, enabling reducing fleet data collection costs.
Lastly, WorldGen-1 is able to make future predictions based on past real data observations, generating multiple potential outcomes from the perspective of the full autonomous driving stack, including pedestrian and vehicle actions, the path of the ego vehicle, and the evolution of the driving scene. This means that WorldGen-1 performs the critical functions of intent prediction and path planning in both simulated and real environments.
With WorldGen-1, we're able to create an infinite number of "digital siblings" of any driving environment, replete with smart agents that have human-like behaviors, from the perspective of the entire AV stack. This takes yet another step toward closing the sim-to-real gap, which is integral to accelerating the development and validation of high end ADAS and L4 autonomous driving systems.
To learn more about WorldGen-1, see the video below and our press release:
https://lnkd.in/g85_4JNi
#generativeai #embodiedai #artifitialintelligence #autonomousdriving #selfdrivingcars #computervision #simulation #deepteaching #helmai
