Reflection AI
1 open Forward Deployed Engineer role at Reflection AI. Curated with comp bands.
On the board since Jul 2026 · still hiring
Our take
Why Reflection AI
America's open-weights frontier bet is staffing a field org before it has shipped a model — pure green-field for people who want to define the FDE function.
- FDE type
- True forward-deployed
- Travel
- Medium
- Comp
- Equity-heavy
- Stage
- Frontier lab
A good fit if you're…
- Builders energized by ground-floor scope at a fast-scaling, pre-launch lab
- Engineers drawn to open-weight and LLM post-training work, not just integration
- Candidates open to a New York, San Francisco or Seoul seat
- People who treat equity and thesis as the bet, not a disclosed cash band
An open-weights frontier lab, pre-launch
Reflection AI, founded in March 2024 by ex-DeepMind researchers Misha Laskin (who led reward modeling for Gemini) and Ioannis Antonoglou (co-creator of AlphaGo), is positioning itself as America's open-frontier lab: training frontier models and releasing the weights, while keeping datasets and training pipelines proprietary — the Llama/Mistral playbook, aimed at enterprises and governments that want ownership and control.
Hiring the field org before the model ships
As of its October 2025 raise the company was about 60 people and had not released a model, yet it has fanned FDE / AI-Engineer postings five deep. That inversion — a field-deployment team staffed ahead of a broad launch — is the defining fact of this seat. You would be defining the function rather than inheriting it, with all the scope and ambiguity that implies.
The post-training FDE variant no one else offers
Among the five postings is an 'LLM post-training' FDE in San Francisco, a flavor unique on the board. Most FDE roles are integration work; this one sits unusually close to the model itself — custom post-training against a customer's data and workflows. If you want forward-deployed work that touches model internals rather than pure deployment plumbing, this is the rare seat that does.
Comp: an equity story, undisclosed base
The listing discloses no band. At a lab this early and this well-capitalized, expect the value weighted toward equity rather than a public top-of-market cash number. Treat the upside as the pitch and pin down the base in-process. A sub-100-person headcount means early grants — but also pre-revenue, pre-product risk that the equity has to carry.
The money and the valuation run-up
Reflection raised $2B at an $8B valuation in October 2025 (backers include Nvidia, Sequoia, Lightspeed, DST, GIC and Eric Schmidt), up roughly 15x from about $545M seven months earlier. By early 2026 it was reportedly in talks to raise again at a $20B-plus valuation. The capital is real; the product, as of those reports, was still unreleased.
Who thrives, who bounces
Thrives: builders energized by ground-floor scope, open-weight and post-training work, and a Seoul, New York or San Francisco seat at a lab racing to its first launch. Bounces: anyone who needs a shipped product to sell, a disclosed comp band, or the stability of an established field playbook. This is a bet on the team and the thesis, not a proven motion.
Questions candidates ask
Has Reflection AI released a model yet?
As of its October 2025 raise and early-2026 reports, no public frontier model had shipped. It plans a first, largely text-based model trained on tens of trillions of tokens, with multimodal capabilities planned later.
What is the 'post-training' FDE variant?
One of the five FDE postings is an LLM post-training role in San Francisco — custom post-training of Reflection's open models for a customer, which sits closer to the model than typical integration-focused FDE work.
Is there really a Seoul seat?
Yes. The postings span New York, San Francisco and Seoul, reflecting a bicoastal-plus-Korea field build-out rather than a single US hub.
What should I ask about comp?
The band isn't disclosed. Pin down base salary, equity grant size, the latest valuation/409A, and the liquidity timeline — at a pre-revenue lab the equity is the entire bet.
Analysis updated 2026-10-29
Open roles at Reflection AI
Compare Reflection AI
Head-to-head
Others in the same tier
Free forever. Unsubscribe anytime.