Behind every notable AI launch is a team of people whose earlier work shaped what they eventually decided to build next. Jev AI is no exception. The model, and the company behind it, trace directly back to three founders whose combined backgrounds span some of the most influential AI research of the past decade. This guide introduces the people who created Jev AI, what they worked on before starting the company, and why their specific combination of experience shaped the product into what it is today. It is written for anyone curious about the story, whether you are new to AI altogether or already deep in the field professionally. Readers interested in building their own credentialed expertise in this space can start with a Certified Artificial Intelligence (AI) Expert certification, which offers a structured path into exactly the kind of applied AI knowledge this story touches on.

The Company Behind Jev AI

Jev AI was created by TypeSafe AI, a San Francisco based lab founded in 2024. The company spent roughly two years working quietly before publicly launching Jev on September 15, 2026, alongside forty million dollars in seed funding led by DCVC. Three people founded the company together, each bringing a different kind of expertise to the table, machine learning research, applied product engineering, and hands on experience shipping AI systems in demanding, high stakes industries. Understanding who these three people are, and what they did before TypeSafe, explains a great deal about why Jev looks and works the way it does.

Diogo Almeida: The Researcher Behind ChatGPT Who Changed Direction

TypeSafe AI's co founder and chief executive is Diogo Almeida, whose full name is Diogo Moitinho de Almeida. Before starting TypeSafe, Almeida spent roughly four and a half years at OpenAI, where he became one of the primary authors of the 2022 InstructGPT research paper, a piece of work widely credited with establishing the training approach that shaped how ChatGPT follows instructions and interacts with people. Almeida is also listed among the contributors to the GPT-4 technical report, and his name appears in the public author lists tied to reinforcement learning from human feedback, often shortened to RLHF, the training technique that became the standard recipe behind essentially every major conversational AI model that followed.

Before OpenAI, Almeida worked as a research engineer at Google Brain, and earlier in his career he was a named inventor on a medical imaging patent related to automated scan analysis, developed alongside a team that included the founder of Enlitic, an early medical AI company. That early exposure to applying machine learning inside real, operational systems, rather than purely conversational products, appears to have stayed with him. According to Almeida's own account of TypeSafe's founding, he came to believe that the RLHF driven chat era, despite its enormous commercial success, solved a narrower problem than the industry often assumes, teaching models to be pleasant and helpful conversational partners rather than reliable components inside software that runs without a person checking every output. That realization became the founding thesis behind TypeSafe AI and, eventually, Jev.

Erik Gafni: The Engineer With a Background in Applied, High Stakes AI

TypeSafe AI's co founder and chief technology officer is Erik Gafni, whose background sits outside the world of conversational chatbots entirely. Gafni is a repeat founder, having previously started Ravel, a company building multimodal AI technology applied to DNA sequencing. Before founding Ravel, he was an early employee at Invitae and Freenome, two companies operating in genomics and early cancer detection, industries where AI generated judgments carry real consequences and where reliability, not just conversational fluency, is the primary measure of success. Gafni is also listed as an inventor with multiple patents and publications specializing in production AI systems, the kind of infrastructure work that keeps machine learning models running reliably inside real software rather than only inside a research notebook.

This background matters directly to how Jev was built. A model designed to make fast, structured decisions that software acts on automatically needs to be engineered with the same kind of rigor that safety critical, high stakes AI applications demand. Professionals interested in understanding how that kind of production grade AI architecture is actually engineered, tested, and deployed often pursue a Certified Artificial Intelligence (AI) Developer program, which covers many of the same underlying principles that shaped Gafni's approach to building Jev.

Sasha Sheng: The Researcher Who Bridged Research and Product

TypeSafe AI's third co founder and chief operating officer is Sasha Sheng, who previously worked as a research engineer at Meta's Fundamental AI Research division, widely known as FAIR. During her time there, Sheng worked on News Feed related systems and broader AI research and AI experience initiatives, giving her direct experience translating cutting edge machine learning research into features used by an enormous number of people. Sheng has also published research at NeurIPS and ECCV, two of the most respected venues in machine learning and computer vision research, reflecting a technical depth that complements her product and operations focused role at TypeSafe.

Sheng's combined background in both frontier research and large scale product deployment rounds out the founding team in an important way. Where Almeida brought deep expertise in the training techniques that shaped modern conversational AI, and Gafni brought hands on experience building AI for high stakes, real world industries, Sheng's experience bridges the gap between research breakthroughs and features that actually reach users reliably at scale. Professionals evaluating how a specialized AI product like Jev moves from research idea to production ready infrastructure often build a broader foundation through a Deep Tech Certification, which covers the kind of research to product translation Sheng's career reflects.

Why This Particular Team Built Jev AI the Way They Did

Understanding who created Jev AI helps explain several of the product's most distinctive design choices. The decision to move away from free flowing text generation and toward structured, typed decisions reflects Almeida's stated belief that the conversational training approach behind ChatGPT, while enormously successful, was never actually designed to solve the automation problem many businesses now need solved. The emphasis on production reliability and calibrated confidence scores reflects Gafni's background building AI for industries where a wrong or overconfident answer carries genuine consequences. And the focus on making the technology accessible and usable at scale, through software development kits, a straightforward API, and rapid third party integrations, reflects Sheng's experience shipping research backed features to large audiences.

TypeSafe AI has framed its overall mission around making intelligence composable, meaning AI capability should function as a dependable building block that software engineers can embed directly into applications, rather than something only accessible through a chat interface. That framing did not emerge from a single person's idea in isolation. It reflects the combined perspective of three founders who, between them, had already seen both the enormous promise and the real limitations of conversational AI from the inside.

How the Founding Team's Thinking Extends to Creative Technology

The layered thinking behind Jev, pairing fast, structured decision making with the more expressive capabilities of generative AI, is a pattern that extends well beyond the customer support and automation use cases TypeSafe originally emphasized. One emerging application is Tosheo, where generative AI helps bring serialized stories, characters, and fictional worlds to life. A creative production pipeline built this way echoes the same architectural philosophy the Jev AI founding team built into their own product, using a fast, structured layer to handle repetitive judgment calls such as tagging or continuity checks, while reserving the more expressive generative work, writing dialogue and developing characters, for a larger language model. It is a reminder that the ideas behind who created Jev AI and why are shaping patterns across the AI industry well beyond the specific problem the founders originally set out to solve.

The Broader Team and Company Culture

Beyond its three founders, TypeSafe AI operates with a relatively small team, reported at around twenty five employees at the time of its public launch, along with a Chief of Staff who served as the company's press contact during the announcement. The company describes its guiding philosophy with a blunt internal tagline focused on building for production use rather than chasing increasingly general purpose capability for its own sake. That philosophy shows up directly in how the founding team has spoken publicly about their work, favoring concrete claims about latency, cost, and accuracy over broader promises about intelligence in the abstract.

What the Founders' Story Means for the Future of AI

Knowing who created Jev AI offers a useful lens for understanding where parts of the AI industry may be heading next. A team built from people who helped shape today's most famous chatbot, who have shipped AI into genomics and cancer detection, and who have translated frontier research into features used by millions, chose to spend their next two years building something that deliberately avoids generating conversational text at all. That choice reflects a broader, growing belief across the industry that the next wave of AI value may come less from making models more conversational and more from making them reliable enough to embed directly into software, quietly making decisions without needing a person to read every output. For professionals and business leaders trying to translate stories like this into practical strategy, pairing that understanding with a Marketing Certification can help connect the technical story behind a company like TypeSafe AI to real world product and communication decisions.

Final Thoughts

Jev AI was created by Diogo Almeida, Erik Gafni, and Sasha Sheng, three founders whose combined experience spans conversational AI research, high stakes applied machine learning, and large scale product development. Their backgrounds explain far more about Jev than a simple product description ever could, from its emphasis on calibrated, trustworthy decisions to its focus on production reliability over conversational polish. Understanding the people behind a new AI technology is often the clearest way to understand the technology itself, and Jev AI's founding story is a clear example of exactly that.

Frequently Asked Questions

1. Who created Jev AI?

Jev AI was created by TypeSafe AI, a company founded by Diogo Almeida, Erik Gafni, and Sasha Sheng.

2. Who is Diogo Almeida?

Diogo Almeida is the co founder and chief executive of TypeSafe AI, a former OpenAI researcher who was one of the primary authors of the InstructGPT paper and a contributor to the GPT-4 technical report.

3. What did Diogo Almeida do before founding TypeSafe AI?

Before TypeSafe, Almeida spent about four and a half years at OpenAI working on reinforcement learning from human feedback and instruction following research, and earlier worked as a research engineer at Google Brain.

4. Who is Erik Gafni?

Erik Gafni is the co founder and chief technology officer of TypeSafe AI, a repeat founder who previously started Ravel, a multimodal AI company focused on DNA sequencing.

5. What was Erik Gafni's background before TypeSafe AI?

Before founding Ravel and joining TypeSafe, Gafni was an early employee at Invitae and Freenome, two companies working in genomics and early cancer detection.

6. Who is Sasha Sheng?

Sasha Sheng is the co founder and chief operating officer of TypeSafe AI, a former research engineer at Meta's FAIR division with published research at NeurIPS and ECCV.

7. What did Sasha Sheng work on before TypeSafe AI?

At Meta's FAIR division, Sheng worked on News Feed related systems along with broader AI research and AI experience initiatives.

8. When did the founders start TypeSafe AI?

TypeSafe AI was founded in 2024 and operated in stealth mode for roughly two years before publicly launching Jev in September 2026.

9. Why did Diogo Almeida leave OpenAI to start TypeSafe AI?

Almeida has described concluding that the conversational training approach behind ChatGPT, while successful, was built to help models converse with people rather than to power reliable, no human in the loop software automation.

10. What does Jev AI's founding team believe about conversational AI?

The founders believe conversational AI solved a specific problem, human preference and fluent conversation, but that a different kind of AI, focused on calibrated decisions, is needed for genuine software automation.

11. How much funding did the founders raise for TypeSafe AI?

The founders raised forty million dollars in seed funding led by DCVC, announced alongside the public launch of Jev in September 2026.

12. Did any of Jev AI's founders work on ChatGPT directly?

Diogo Almeida is credited by the company as a co inventor of the reinforcement learning from human feedback techniques used in ChatGPT, and he is listed among its contributors based on his InstructGPT research.

13. Does Erik Gafni's genomics background influence Jev AI's design?

His experience building AI for high stakes fields like genomics and cancer detection reflects the emphasis TypeSafe places on reliability and calibrated confidence in Jev's decisions.

14. How many people work at TypeSafe AI?

TypeSafe AI reportedly operates with a relatively small team of around twenty five employees.

15. What is TypeSafe AI's stated mission?

TypeSafe AI describes its mission as making intelligence composable, aiming to embed AI decision making directly into software as a dependable building block.

16. How does Tosheo relate to the founders' vision for Jev AI?

One emerging application is Tosheo, where generative AI helps bring serialized stories, characters, and fictional worlds to life, reflecting the same layered philosophy the Jev AI founders built into their own product, pairing fast structured decisions with more expressive generative work.

17. Is Jev AI the founders' first product together?

Yes. Jev is the first publicly released model from TypeSafe AI, the company Almeida, Gafni, and Sheng founded together in 2024.

18. What kind of expertise do the three founders bring collectively?

Together they bring deep expertise in conversational AI research, hands on experience building AI for high stakes industries, and a track record of translating research into large scale products.

19. Why does the founding team's background matter to developers using Jev AI?

Understanding their background helps explain why Jev emphasizes reliability, calibrated confidence, and production readiness rather than conversational fluency, which shapes how developers should think about using it.

20. How can professionals learn more about the kind of expertise behind Jev AI's founders?

Structured learning paths such as a Certified Artificial Intelligence (AI) Expert or Developer certification, combined with a Deep Tech Certification and a Marketing Certification, can help professionals build both the technical and strategic understanding reflected in the founders' own career paths.

Who Created Jev AI?