Untapped Ventures Invests in Lemurian Labs

Investing in Lemurian Labs to Commercialize a Compute Paradigm that Reduces AI Costs

Jay Dawani and Dr. Vassil Dimitrov are no strangers to going great lengths to solve some of the world’s most challenging problems.

Dawani comes with an impressive background in deep learning, robotics, and artificial intelligence (AI). In fact, he authored Mathematics for Deep Learning and is an advisor to NASA’s Frontier Development Lab. Before co-founding Lemurian Labs, Dawani sold his previous AI company and was previously a qualifier of the 2012 Olympics and vice-captain of the Pakistan national swimming team. The same tenacity that comes with training for the Olympics is what drove Dawani to supercharge the next generation of AI processing technology. 

He is joined by co-founder Dr. Vassil Dimitrov, a professor with over 20+ years of experience in applied mathematics. Dimitrov’s discoveries in processing and image compression technology have revolutionized modern healthcare - from modern hearing aids and Pillcam systems. 

Together, Dawani and Dimitrov want to change the way AI technology reaches the hands of billions of consumers. 

AI’s Big Constraint

In an ever-evolving landscape of AI technology, one crucial bottleneck remains: the limitations of current computing architectures that enable AI technologies.

At Lemurian Labs, an early-stage startup from Google, Intel, and Nvidia alum, Dawani and Dimitrov uncovered a game-changing solution. Lemurian Labs is on a mission to democratize AI, and through the development and deployment of novel mathematical formulas, address several of the key challenges currently facing the industry. Their next generation Spatial Processing Unit (SPU) tackles the inefficiencies of traditional GPUs head-on. 

In one small microchip, Dawani and Dimitrov are ushering in a new era of AI-enabled applications, from autonomous vehicles to healthcare robots. The Lemurian Labs SPU is an innovative AI inference accelerator that enables the next generation of AI workloads for data centers and enterprise companies. 

Creating Energy Efficiency

Lemurian Labs envisions a future where AI seamlessly integrates into everyday life, which requires the cost and power consumption of AI to trend down.  Unfortunately, the inverse is true where escalating costs and energy consumption of AI inference, particularly as its adoption grows, are unsustainable. 

For example, deploying GPT3 at the scale of Google search would consume 400 megawatts, enough electricity to power up to ~360K conventionally-powered homes for an entire year, with costs exceeding $100B. Not to mention, the type of GPU needed by large enterprises would drastically leave a large carbon footprint. 

Data centers currently consume 1.3% of the world’s total energy, and this is expected to quadruple by 2030 with the growth in AI. Rising AI inference costs and energy use are unsustainable, especially as more companies leverage this technology. Current solutions do not meet the demand by consumers and large enterprises innovating at scale with AI products.

Lemurian’s Solution

“AI models have grown in complexity and adoption at a rate so rapid that it is now stressing our current hardware infrastructure to the point of fracture,” said Jay Dawani, CEO and co-founder of Lemurian Labs. “Models will continue to grow so we need fundamental changes in hardware in order to keep pace. We reimagined accelerated computing around the needs of AI models and its developers to bring down the cost and power consumption by 10X. This way we can fundamentally change the economics of AI such that any company can train and deploy LLMs, not just the fortunate few.”

Lemurian Labs developed an accelerated computing platform that leverages a novel number system, referred to as PAL (parallel adaptive logarithm), to expedite AI workloads by up to 20x greater throughput at a 1/10th of the total cost of legacy GPUs. This combination of a new number system and co-designed hardware and software in their compute platform yields performance gains, without leaving a large carbon footprint, allowing users to deploy more within the same power envelope. 

With faster AI workload deployments at a lower cost, Lemurian Labs is accelerating developer productivity while making these solutions accessible to a wider range of businesses and organizations.

At Untapped Ventures, we recognize that the complexities of AI technologies demand simple solutions developed for the future, and Lemurian Labs is at the forefront of this transformative shift. 

Investment

We are proud to support Lemurian Labs by backing their latest $9 million round of seed funding as reported by TechCrunch, alongside Oval Park Capital, Good Growth Capital, Raptor Group, Alumni Ventures, and others.

Together, we are thrilled to partner with Jay Dawani and Dr. Vassil Dimitrov, whose expertise and commitment to revolutionizing AI align seamlessly with our vision of elevating humanity – with sustainable AI hardware platforms that power a more energy-efficient future.