Diogo Almeida
Founder and CEO of TypeSafe AI; co-author of OpenAI's InstructGPT paper on training with human feedback
- Founder and CEO, TypeSafe AI12
Diogo Almeida is the founder and CEO of TypeSafe AI, a San Francisco company that builds what it calls System One models, which return typed, probabilistic decisions rather than generated text. Almeida is a co-author of "Training language models to follow instructions with human feedback" (2022), the OpenAI paper that introduced InstructGPT, and TypeSafe's team page notes earlier work at Google Brain. In September 2026 Almeida announced TypeSafe's first public model, Jev, together with the company's published workflow evaluations.142
Career
Almeida's past roles include work at Google Brain and at OpenAI. In TypeSafe's launch post, Almeida writes of helping build, at OpenAI, the methods that made language models follow instructions and talk with people, and Almeida is listed among the authors of the March 2022 InstructGPT paper.124
Almeida founded TypeSafe AI and leads it as CEO, alongside COO Sasha Sheng and CTO Erik Gafni. The launch post says the company spent two years in stealth before its public debut.12
TypeSafe AI and Jev
On September 15, 2026, Almeida announced Jev, TypeSafe's first System One model, in early access. The post describes a new model architecture, a parallel sampler and a training method the company calls Reinforcement Learning for Calibrated Decisions (RLCD), with Jev returning structured decisions instead of free-form strings.2
The same post introduced TypeSafe's workflow evaluations, which break automation tasks such as security incident triage, invoice processing and customer service into narrow typed questions and compare models against consensus labels.23
Work
WorkflowEvals (TypeSafe AI)Open-source code
Wrote TypeSafe's launch post, which introduced the company's workflow evaluations and pointed readers to the workflow evals site with its examples, queries and workflows.23
InstructGPT (external site: arxiv.org)Published research
Co-author of the 2022 paper that fine-tuned GPT-3 with supervised demonstrations and reinforcement learning from human feedback to produce the InstructGPT models.4
Career timeline
Other documented roles
Roles the sources document but that are not confirmed as current.
- Researcher on instruction-following language models (past), OpenAI24
- Past role listed on TypeSafe's team page, Google Brain1
Related people
More sourced profiles at the same organization, in the same field, or from the same school.
- Erik GafniAlso at TypeSafe AI
- Aylin CaliskanSafety and alignment
- Sasha ShengAlso at TypeSafe AI
- Collin BurnsAlso at OpenAI
- Abhimanyu DasAlso at Google
- Eric Michael SmithSafety and alignment
- Adam PaszkeAlso at Google
- Ethan PerezSafety and alignment
- Aditya RameshAlso at OpenAI
- Ian R. McKenzieSafety and alignment
- Aidan GomezAlso at Google Brain
- Margaret MitchellAlso at Google AI
Is this you?
Profiles cover public professional work only, and every line cites a source. If something is out of date, missing or wrong, send a correction with a link to an official page and we will update it. Suggest a correction
Add this badge to your website or GitHub profile README. It links back to this page. It is not an endorsement or a rating.
Markdown (for a README)
[](https://unitedstatesofamericasuperintelligence.com/people/diogo-almeida/)
HTML
<a href="https://unitedstatesofamericasuperintelligence.com/people/diogo-almeida/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/profile.svg" alt="Diogo Almeida — Profile on USASI" height="20"></a>
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.