# Tulsee Doshi

Markdown copy of https://realalpha.app/builders/tulsee-doshi/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

Real Alpha has cut 8 moments of Tulsee Doshi (Senior director & head of product for Gemini, Google DeepMind) talking, from an interview on Lenny's Podcast published Nov 8, 2026.

- Role: Senior director & head of product for Gemini, Google DeepMind
- About: Leads the Gemini product roadmap and strategy after five years heading Google's Responsible AI product work.
- Company: [Google DeepMind](https://realalpha.app/companies/#google-deepmind)
- On X: https://x.com/tulseedoshi

## What Tulsee Doshi said

- [Model PMs have to discover what training produces](https://realalpha.app/s/model-pms-have-to-discover-what-training-produces/)
  Lenny's Podcast · Nov 8, 2026 · 3 min
  > “the model isn't predictable. And so it's kind of like, like Snape's potions class, right?” — Tulsee Doshi · [3:13](https://realalpha.app/v/qGf2Yvan4Wg/?t=193)
- [Build for what models can do now and next](https://realalpha.app/s/build-for-what-models-can-do-now-and-next/)
  Lenny's Podcast · Nov 8, 2026 · 2 min
  Doshi’s 60-20-20 framework balances working features with capabilities that are nearly ready and more ambitious ideas. It lets a product work today while improving as models advance.
- [Better evaluations can expose gaps in representation](https://realalpha.app/s/better-evaluations-can-expose-gaps-in-representation/)
  Lenny's Podcast · Nov 8, 2026 · 3 min
  > “Realistically, six skin tones cannot capture the diversity of skin tones in the world.” — Tulsee Doshi · [27:19](https://realalpha.app/v/qGf2Yvan4Wg/?t=1639)
- [Google's safety plan combines monitoring with cyber defense](https://realalpha.app/s/googles-safety-plan-combines-monitoring-with-cyber-defense/)
  Lenny's Podcast · Nov 8, 2026 · 3 min
  > “deploying monitors that can monitor the model's chain of thought and actually identify whether the model is misaligned with its directed actions” — Tulsee Doshi · [11:06](https://realalpha.app/v/qGf2Yvan4Wg/?t=666)
- [Research PMs should test models and start with small evals](https://realalpha.app/s/research-pms-should-test-models-and-start-with-small-evals/)
  Lenny's Podcast · Nov 8, 2026 · 3 min
  > “start with even just five prompts, right? Like start, you don't have to start large.” — Tulsee Doshi · [23:29](https://realalpha.app/v/qGf2Yvan4Wg/?t=1409)
- [Teams need time to build and share AI workflows](https://realalpha.app/s/teams-need-time-to-build-and-share-ai-workflows/)
  Lenny's Podcast · Nov 8, 2026 · 5 min
  > “it is hard to create the space to actually take a step back and make sure you know the tools well.” — Tulsee Doshi · [39:24](https://realalpha.app/v/qGf2Yvan4Wg/?t=2364)
- [Model behavior is part of the product's UX](https://realalpha.app/s/model-behavior-is-part-of-the-products-ux/)
  Lenny's Podcast · Nov 8, 2026 · 4 min
  > “it's also important to think about the model as having a UX also.” — Tulsee Doshi · [17:02](https://realalpha.app/v/qGf2Yvan4Wg/?t=1022)
- [Google pairs frontier models with a full-stack strategy](https://realalpha.app/s/google-pairs-frontier-models-with-a-full-stack-strategy/)
  Lenny's Podcast · Nov 8, 2026 · 3 min
  > “we actually go all the way from the hardware to then the modeling to then this kind of end user application.” — Tulsee Doshi · [7:41](https://realalpha.app/v/qGf2Yvan4Wg/?t=461)

## The episode

- [Why Google wants to win the frontier conversation | Tulsee Doshi (Head of Product for Gemini)](https://realalpha.app/v/qGf2Yvan4Wg/) · Lenny's Podcast · Nov 8, 2026