×
TypeSafe AI
System One Models: typed, calibrated decisions for software, not chat
Last verified September 17, 2026 · Updated daily
What TypeSafe AI is building
TypeSafe is a lab that took the opposite research direction from chat models. Its first model, Jev, does not produce strings. It produces typed values: a yes/no probability, a selection from a defined list, a score on a scale you specify, each with a calibrated confidence attached. The pitch is that a model whose output is a type can be wired into software the way a function is, with no parsing, no prompt-injection-shaped string handling, and no type errors. The company calls this class a System One Model, borrowing Kahneman's fast-intuitive-judgment framing. The training method is what they call Reinforcement Learning for Calibrated Decisions (RLCD): instead of RLHF's objective of pleasing a human rater, the reward is epistemic honesty about probabilities on decision tasks. Diogo Almeida co-invented RLHF at OpenAI, so the framing is deliberate: 'We've been optimizing for humans, and we're superhuman at pleasing humans.' Jev is his attempt to optimise for the other consumer of intelligence, which is code. Architecturally, Jev generates all outputs in a single query rather than sequentially decoding tokens, which is where the latency claim comes from: the company quotes 70 to 500 ms per decision versus 3 to 329 seconds for frontier LLMs on the same tasks, and prices input at $0.042 per million tokens with output tokens free. The target workloads are 'smart if-statements' inside pipelines: insurance underwriting triage, service-request classification, invoice evaluation, security-alert triage, and grading the outputs of other agents. Early access opened in September off a waitlist.
Why this matters
The unit economics of running an LLM as a classifier are bad and everyone building agent pipelines knows it: you pay for a full autoregressive decode to get back the word 'yes'. TypeSafe's own workflow benchmarks claim 193.6x faster and 444.6x cheaper than comparable LLM calls on representative production tasks; those are company numbers, not independent ones, but even a fraction of that gap changes what you can afford to put a model in front of. Dealroom put this round in the 99th percentile of all AI seed rounds ever by size (sample of 28,473 deals), and DCVC's James Hardiman framed the bet as reliable embedded models being 'one of the biggest remaining challenges in AI'. If calibrated, typed decisions become a primitive, the verification and guardrail layer of every agent stack becomes a customer.
Open roles at TypeSafe AI
6 positions we're tracking. Roles are re-checked daily and removed when filled.
Member of Technical Staff, Model Capabilities (datasets, evals, data tooling; Python)
First seen today
Member of Technical Staff, Backend/Platform (inference API, multi-cloud serving)
First seen today
Member of Technical Staff, Infrastructure (Kubernetes, GPU autoscaling, Pulumi)
First seen today
Developer Advocate (docs, SDK references, Discord, early-access programs)
First seen today
Founding Marketer (first marketing hire, narrative and category)
First seen today
Member of Staff (self-scoped role, 5+ years experience)
First seen today
Know when TypeSafe AI is hiring before anyone else
A role stays uncontested for about four days. Here's the window — and where we put you in it.
From $5.99/week, cancel any time.
Watching TypeSafe AI
0 applicantsRole spotted & verified
1You get the alert
1You've applied
~8Hits the job boards
250+Hiring outlook
Six live roles on the company's Ashby board as of 17 Sep 2026, all San Francisco office: three Member of Technical Staff openings (Model Capabilities, Backend/Platform, Infrastructure/Kubernetes), a Developer Advocate, a Founding Marketer, and an open-ended 'Member of Staff' post. The job posts describe a small team from OpenAI, Google Brain and Meta/FAIR building the inference API before it ships publicly.
Working at TypeSafe AI
TypeSafe AI is a AI company based in San Francisco, USA. Working at a AI company at this stage means opportunity to shape your role based on the company stage.
San Francisco, USA is where most TypeSafe AI positions are located.
How to actually get hired at TypeSafe AI
Why applying the normal way doesn't work
TypeSafe AI runs an applicant tracking system, but hiring managers still work referrals first. A cold application to TypeSafe AI isn't dead, it's just fourth in line behind internal referrals, sourced candidates and recruiter pipelines.
Who to contact at TypeSafe AI
What to show them
Write a cookbook that replaces an LLM-based classification step in a real open-source pipeline with a typed-decision call, with a before/after on latency and cost per 1,000 workflows. Publish it; that is the content this role owns.
A cold email that works at TypeSafe AI
What TypeSafe AI screens for
DCVC writes deep-tech cheques and does not usually lead $40M seeds; this is a conviction bet on a research thesis, not a product with revenue. The hiring board tells you what phase the company is in: the Backend/Platform post says the API has 'uptime, reliability and backwards compatibility requirements (once we ship)', so the serving layer is being built right now, ahead of general availability. The Model Capabilities post calls evaluation tooling the 'secret sauce', which means eval and dataset engineers get unusual leverage here. Two years of stealth plus a public launch in the same week as the raise means the next 90 days set the founding team; the Founding Marketer and Developer Advocate roles are literally first hires in their functions.
Customize your CV for the TypeSafe AI role. Matching the job description language helps clear ATS filters.
Don't make these mistakes
Diogo co-invented RLHF and InstructGPT and left OpenAI to build the thing he thinks RLHF got wrong. Do not open with how excited you are about LLMs or agents; the whole company is a critique of chat-shaped models. Do not pitch 'structured output' as if it were novel to him: he knows JSON mode exists and Jev is an argument that it is the wrong abstraction. Skip the 'ChatGPT co-creator' flattery entirely, it is in every press piece. Lead with a concrete case where an LLM classifier in your pipeline was miscalibrated, slow, or expensive enough that you ripped it out, and what you replaced it with.
Mistakes that kill TypeSafe AI applications
A recycled CV gets rejected fast at TypeSafe AI ( people). They notice.
Don't open with what you want. Open with what TypeSafe AI is dealing with right now — The unit economics of running an LLM as a classifier are bad and everyone building agent pipelines knows it: you pay for a full autoregressive decode to get back the word 'yes', and what you'd do about it.
Most AI applications get ghosted. A day-five follow-up can double your response rate.
Applying to TypeSafe AI? Get the contact, not the form.
The TypeSafe AI interview process
4 stages · 14 days typical · take-home: yes · modelled from similar companies
We don't yet have verified candidate reports for TypeSafe AI. What follows is the typical process for a -person AI company — treat it as a model, not confirmed detail.
Interview stages
Recruiter Screen
Phone or video · 30 min
Basic qualification and logistics
Recruiter or HR
Hiring Manager Interview
Video call · 45 min
Role fit and experience deep-dive
Hiring manager
Technical/Functional Round
Video call · 60 min
Skills assessment and problem-solving
Team members
Final Round
In-person or video · 60 min
Culture fit and cross-functional alignment
Senior leadership
TypeSafe AI take-home assignment
TypeSafe AI includes a take-home exercise in their interview process. For AI roles, this typically involves a practical problem that takes 2-4 hours. Focus on clean, working code over premature optimization. They're evaluating how you think and communicate, not just the solution.
TypeSafe AI interview timeline
Expect days total. Compared to similar AI companies (14 days median), TypeSafe AI is about average.
Interviewed at TypeSafe AI?
Tell us how it went — stages, questions, timeline. Takes 90 seconds and it's how this page stays accurate for the next person.
Submit your TypeSafe AI interview experience →TypeSafe AI jobs, frequently asked questions
How many jobs does TypeSafe AI have open?
As of September 2026, TypeSafe AI has 6 open positions.
Does TypeSafe AI hire remotely?
All current TypeSafe AI roles are based in San Francisco, USA.
What roles is TypeSafe AI hiring for?
TypeSafe AI is hiring across Data, Engineering, Other, Marketing. The most recent opening is Member of Technical Staff, Model Capabilities (datasets, evals, data tooling; Python).
How do I apply for a job at TypeSafe AI?
Use the apply links above, or check our guide to getting hired at TypeSafe AI.
Does TypeSafe AI respond to cold emails?
Response rate data for TypeSafe AI not yet confirmed.
Who is the hiring manager at TypeSafe AI?
At this size, hiring is usually run by the hiring manager.
How competitive is it to get hired at TypeSafe AI?
Expect 100-250 applicants in the first two weeks for AI roles at this size. Apply within 72 hours for best odds.
How many rounds is the TypeSafe AI interview?
4 stages: Recruiter Screen, Hiring Manager Interview, Technical/Functional Round, Final Round.
Is the TypeSafe AI interview hard?
Expect technical depth and system design, not algorithm trivia. Candidates report Technical Interview as the toughest stage.
Does TypeSafe AI give a take-home task?
Yes, TypeSafe AI includes a take-home assignment.
How long does TypeSafe AI take to get back to you?
Around 14 days across the full process.
What should I prepare for the TypeSafe AI interview?
technical depth and system design is the priority. Show you can work autonomously — that matters more than algorithms at people.
Where is TypeSafe AI based?
TypeSafe AI is headquartered in San Francisco, USA.
Get TypeSafe AI roles before they're posted
A role stays uncontested for about four days. Here's the window — and where we put you in it.
From $5.99/week, cancel any time.
Watching TypeSafe AI
0 applicantsRole spotted & verified
1You get the alert
1You've applied
~8Hits the job boards
250+Related
More AI companies hiring
Other startups we track in the same space.