Transluce
1 open Forward Deployed Engineer role at Transluce. Curated with comp bands.
Comp on the board: $310K–$500K.
On the board since Jul 2026 · still hiring
Our take
Why Transluce
Frontier-lab money ($310K–$500K) at a nonprofit eval lab whose FDE embeds with regulators and frontier labs — not enterprises — to audit how models actually behave.
- FDE type
- True forward-deployed
- Travel
- Medium
- Comp
- Top of market
- Stage
- Early-stage
A good fit if you're…
- Engineers who want mission and frontier-lab comp in the same seat
- People drawn to AI evaluation, interpretability, and model oversight
- Builders who can turn policy needs into eval harnesses and shipped deliverables
- Candidates fine with a nonprofit's governance, and with regulators and labs as the customer
A nonprofit founded to understand models
Transluce is the nonprofit AI-evaluation lab co-founded in October 2024 by Jacob Steinhardt (UC Berkeley) and Sarah Schwettmann (MIT), building open-source tooling to understand AI systems and steer them in the public interest. Its research spans investigator agents that surface harmful model behaviors, circuit and user-modeling analysis, and behavioral evaluation — the substrate the forward-deployed seat works on top of.
What the FDE actually does
The AI Behavior Engineer executes Transluce's government-evaluator contracts — including building harmful-manipulation evaluations with the EU AI Office — designs privileged-access oversight exercises with frontier labs, and adapts behavioral-evaluation pipelines to partner contexts like mental health and persuasion. Deployment here means embedding with policymakers and civil-society partners, not paying enterprise customers.
Docent: the tooling you'd deploy
Docent, Transluce's system for analyzing and intervening on AI agent behavior, was introduced in March 2025, reached public alpha in August 2025, and was open-sourced under Apache 2.0 in September 2025. An FDE stands up eval harnesses on tooling like this and translates the findings into oversight that regulators and labs can act on.
Comp: nonprofit pay that isn't
$310K–$500K plus benefits is the plot twist. Nonprofit status usually implies a discount; instead the band matches frontier commercial labs. That tells you Transluce is competing head-on for the same evaluation talent OpenAI and Anthropic want — and prices accordingly rather than trading cash for mission.
Why regulators and labs both take the calls
Founder pedigree (Berkeley and MIT), open-source releases like Docent, and privileged access inside frontier labs give a small organization unusual leverage: it sits between the labs building the models and the regulators overseeing them. For an FDE, that means the work lands in real oversight decisions, not slideware.
Who thrives, and who bounces
Thrives: engineers energized by public-interest oversight, comfortable with ambiguity and stakeholder-facing work. Bounces: anyone who wants a clear enterprise product roadmap or a conventional paying customer — here the customer is the public-interest ecosystem, and the deliverable is an eval or a finding, not a renewal.
Questions candidates ask
Is Transluce really a nonprofit paying $310K–$500K?
Yes. The AI Behavior Engineer listing posts $310,000–$500,000/year plus benefits — unusual for a nonprofit and on par with frontier commercial labs.
Who are Transluce's customers?
Government evaluators such as the EU AI Office, civil-society partners, and frontier labs — not enterprise buyers. The mission is public-interest oversight of AI systems.
What would I build day to day?
Behavioral-evaluation pipelines and eval harnesses (often on tooling like Docent), privileged-access oversight exercises with labs, and evaluations for things like harmful manipulation and persuasion.
Do I need interpretability research experience?
It helps — the work draws on interpretability and behavioral evaluation — but the FDE seat is applied: translating policy needs into shipped evals and findings that partners can use.
Analysis updated 2026-09-25
Open roles at Transluce
Compare Transluce
Others in the same tier
Free forever. Unsubscribe anytime.