About
AI is changing how engineering teams are built and run, and the companies that adapt fastest will pull ahead. The best engineering leaders will combine traditional management with their own hands-on agentic engineering.
I've built and managed engineering organizations for over a decade — finding great engineers, creating a culture where they do their best work, and scaling teams that consistently ship.
At Nervana, I founded the software team from scratch and grew it to 10 engineers. After the Intel acquisition, I scaled the team to 150 across the US, Poland, India, and Israel — without losing the quality and culture that made us worth acquiring. As co-founder of Luminide, I led product, engineering, and go-to-market — through acquisition by Akridata. Most recently, as CTO at ThirdLaw, I built runtime safety and observability for LLM and agent behavior.
Underneath it all: 20+ years of hands-on engineering and research, a Ph.D. in CS, and published work at NeurIPS. I know how engineers think because I am one — these days, an agentic one.
Experience
- CTO, ThirdLaw 2025–2026Runtime AI safety: observability and governance for LLM and agent behavior in the enterprise.
- VP of Engineering, TensorWave 2024Engineering, recruiting, and datacenter/SaaS operations.
- AI Engineering & Leadership, consulting 2023–2025Advised companies and investors; engagements including DE Shaw and GLG.
- Senior Director, Product, Akridata 2022–2023Integrated Luminide's IDE; led GTM and community (Kaggle, ML Commons).
- Co-founder & CEO, Luminide 2020–2022AI tools for model accuracy and lower training cost; acquired by Akridata.
- Director of AI Software, Intel 2016–2019Grew the team from 10 to 150 across four countries; three AI chip bring-ups; co-developed Flexpoint.
- Director of AI Software, Nervana 2014–2016Founded the software team for the first AI hardware accelerator; acquired by Intel.
Publications & Patents
- Flexpoint: An Adaptive Numerical Format for Efficient Training of Deep Neural Networks — NeurIPS, 2017.
- Dynamic Management of Numerical Representation in a Distributed Matrix Processor Architecture — U.S. Patent 10,552,119.
- Expedited Assessment and Ranking of Model Quality in Machine Learning — U.S. Patent Application 17/560,422.
- Real-time Analytics: Techniques to Analyze and Visualize Streaming Data — Byron Ellis, edited by Luke Hornof. Wiley, 2014.
- Compiling for Template-based Run-time Code Generation — Journal of Functional Programming 13(3), 2003.
Talks
Speaking on engineering leadership and building in the AI era, at SF AI meetups and founder events. Recent topics:
- Building Amazing Engineering Teams (AI-era update)
- Bootstrapping a Startup — the Luminide story
- Company Models — the AI OS for Companies
- Where the High-Potential Startup Areas Are
- The Times They Are AI-Changin' — startup essentials
Contact
SF Bay Area. Email luke@hornof.org, or find me on the profiles linked in the sidebar.