# Drishan Arora

> Drishan Arora is the co-founder and CEO of Deep Cogito, a San Francisco AI lab building toward general superintelligence through openly released reasoning models. A former senior engineer at Google who led large language model work for its generative search, Arora left to chase a contrarian bet: that intelligence improves not by searching harder but by building better intuition. His team trained the Cogito family of hybrid reasoning models - which can answer instantly or stop to think - in roughly 75 days, then scaled to a 671-billion-parameter model his company calls the strongest open-weight LLM from a US firm. The method underneath it all, iterated distillation and amplification, lets the models internalize their own reasoning paths and self-improve.

- **Role:** Co-Founder & CEO at Deep Cogito
- **Organizations:** Deep Cogito, Google (former)
- **Nationality:** Indian
- **Education:** B.Tech, Electrical Engineering, IIT Delhi, Master's, Computer Science, Columbia University
- **Known for:** Co-founded and leads Deep Cogito, an AI lab building toward general superintelligence, Shipped the Cogito family of open hybrid reasoning models that topped open-model charts on release, Scaled to a 671B-parameter mixture-of-experts model billed as the strongest open-weight LLM from a US company

## Career timeline

- **2014-2016** — Master's in Computer Science at Columbia University; teaching assistant for five courses
- **post-2016** — Senior Software Engineer at Google; tech lead for question answering across Google Search and Google Assistant, Featured Snippets team
- **pre-2024** — Led large language model modeling for Google's generative search product
- **2024-06** — Co-founded Deep Cogito in San Francisco with Dhruv Malhotra
- **2025-04** — Deep Cogito emerges from stealth with Cogito v1 hybrid reasoning models (3B-70B), trained in ~75 days
- **2025-07** — Releases Cogito v2 preview - four open models up to 671B with self-improving intuition via IDA
- **2025-08** — $13M seed round led by Benchmark; South Park Commons among backers
- **2025-11** — Releases Cogito v2.1 671B, described as the best open-weight LLM by a US company

## Achievements

- Co-founded and leads Deep Cogito, an AI lab building toward general superintelligence
- Shipped the Cogito family of open hybrid reasoning models that topped open-model charts on release
- Scaled to a 671B-parameter mixture-of-experts model billed as the strongest open-weight LLM from a US company
- Pioneered productized iterated distillation and amplification (IDA) for model self-improvement
- Trained competitive frontier-class open models in roughly 75 days with a small team
- Previously led question answering systems behind Google Search and Google Assistant

## Latest updates

- **2025-11** — Announced Cogito v2.1 671B, calling it the best open-weight LLM by a US company, competitive with frontier closed and open models.
- **2025-08** — Closed a $13M seed round led by Benchmark.
- **2025-07** — Released the Cogito v2 preview - four open hybrid reasoning models up to 671B parameters built on the IDA method.

## Links

- Website: https://deepcogito.com
- LinkedIn: https://www.linkedin.com/in/drishanarora/
- Twitter/X: https://x.com/drishanarora

---

Profile page: https://yespress.io/drishan-arora
Published by YesPress — https://yespress.io
Last updated: 2026-06-24