# Stefano Ermon

> Stefano Ermon is the CEO and co-founder of Inception Labs, the company behind Mercury - the world's first commercial-scale diffusion LLM. An Italian-born Stanford associate professor on leave, Ermon co-invented DDIM (the speedup trick powering modern image generation), Direct Preference Optimization (DPO, the alignment technique used to fine-tune today's leading LLMs), and the SEDD discrete diffusion framework that won ICML 2024's Best Paper Award. At Inception, he's betting that diffusion - not autoregression - is the architecture that will define the next generation of AI, with Mercury 2 clocking 1,000+ tokens per second at a fraction of the cost of GPT or Claude.

- **Role:** Chief Executive Officer at Inception Labs
- **Organizations:** Inception Labs, Stanford University, Woods Institute for the Environment
- **From:** Northeast Italy (Alpine region)
- **Nationality:** Italian
- **Education:** Bachelor of Science in Electrical and Electronics Engineering, Università degli Studi di Padova, Master of Science in Electrical and Electronics Engineering, Università degli Studi di Padova, Master of Science in Computer Science, Cornell University, PhD in Computer Science, Cornell University
- **Known for:** Co-invented DDIM (Denoising Diffusion Implicit Models), enabling 10-50x faster sampling in modern diffusion models, Co-invented Direct Preference Optimization (DPO), now a standard technique for aligning large language models, Co-invented FlashAttention, a widely-used memory-efficient attention mechanism

## Career timeline

- **2015** — Joined Stanford University as faculty in Computer Science
- **2017** — Awarded Sloan Research Fellowship and NSF CAREER Award
- **2018** — Received IJCAI Computers and Thought Award - premier award for AI researchers under 35
- **2020** — Co-authored DDIM (Denoising Diffusion Implicit Models), enabling 10-50x faster diffusion sampling
- **2021** — DDIM wins Outstanding Paper Award at ICLR 2021
- **2022** — Co-authored DPO (Direct Preference Optimization), a foundational LLM alignment technique
- **2023** — DPO wins Outstanding Paper Runner-up at NeurIPS 2023; co-authored SEDD discrete diffusion framework
- **2024** — SEDD wins Best Paper Award at ICML 2024 (1 of 10 among 9,473 submissions); co-founds Inception Labs
- **2025** — Inception Labs raises $50M Series A led by Menlo Ventures; Mercury LLM launched from stealth
- **2025** — Released monograph 'The Principles of Diffusion Models' with co-authors including Yang Song
- **2026** — Mercury 2 launched - world's first reasoning diffusion LLM, achieving 1,000+ tokens per second

## Achievements

- Co-invented DDIM (Denoising Diffusion Implicit Models), enabling 10-50x faster sampling in modern diffusion models
- Co-invented Direct Preference Optimization (DPO), now a standard technique for aligning large language models
- Co-invented FlashAttention, a widely-used memory-efficient attention mechanism
- ICML 2024 Best Paper Award for 'Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution'
- ICLR 2021 Outstanding Paper Award for DDIM
- ICLR 2022 Outstanding Paper Award
- IJCAI Computers and Thought Award (2018) - premier AI award for researchers under 35
- Sloan Research Fellowship (2017)
- NSF CAREER Award, ONR Young Investigator Award, AFOSR Young Investigator Award
- H-Index of 85 on Semantic Scholar with 62,708+ citations
- Built and leads Inception Labs, which raised $50M with backers including Andrew Ng and Andrej Karpathy
- Launched Mercury 2 - the world's first reasoning diffusion LLM at 1,000+ tokens per second

## Latest updates

- **2026-02** — Launched Mercury 2, the world's first reasoning diffusion LLM, achieving 1,000+ tokens per second - 5x faster than leading speed-optimized models
- **2025-11** — Inception Labs raised $50M Series A led by Menlo Ventures, with participation from NVIDIA, Microsoft M12, Snowflake Ventures, Databricks, Andrew Ng, and Andrej Karpathy
- **2025-10** — Released monograph 'The Principles of Diffusion Models' with collaborators including Dr. Yang Song
- **2025-02** — Inception Labs emerged from stealth with Mercury, the world's first commercial-scale diffusion LLM
- **2024-01** — Co-founded Inception Labs in Palo Alto, California

## Links

- Website: https://cs.stanford.edu/~ermon/
- LinkedIn: https://www.linkedin.com/in/ermon/
- Twitter/X: https://x.com/StefanoErmon
- GitHub: https://github.com/ermongroup

---

Profile page: https://yespress.io/stefano-ermon
Published by YesPress — https://yespress.io
Last updated: 2026-05-13