RL Environment Engineer

Sunset

New York (NY)

On-site

USD 220,000 - 320,000

Full time

5 days ago

Be an early applicant

A complete application in a minute

Tailored resume and cover letter, ready to send.

Get past ATS filters

Benefits offered by this job

Meaningful equity

Medical, dental & vision

Unlimited PTO

Gym memberships

Desk allowance

Relocation bonus

Job summary

Replay is seeking an RL Environment Engineer to turn real enterprise data into production‑grade training environments for frontier AI labs. You will design tasks, tools, and evaluators, partnering with researchers to close capability gaps and ensure reproducible, safe rollout of environments.

The role demands shipping ownership, strong software engineering, and experience building environments used for real training/evaluation, with a focus on robustness and measurable signal quality.

Qualifications

3-5+ years of ML / software engineering experience shipping production systems.
Built production‑grade RL environments that were used in real training or evaluation, not just prototypes.
Ability to turn ambiguous goals from researchers into well‑scoped, working systems.
A clear grasp of environment design tradeoffs: task specification, state and tool design, determinism, difficulty calibration and solvability.
Built evaluation or agentic harnesses, with rigor: precision, recall, calibration, distribution shift.
High ownership, action‑oriented mindset: take problems from idea to shipped.
Startup experience and comfort with evolving requirements.
Clear written and verbal communication with researchers, engineers, and non‑technical partners.]
job_responsibilities_description ["Design and build production‑grade RL environments grounded in real enterprise workflows, from task spec through to a stable, versioned release","Write reward functions and programmatic verifiers that score outcomes reliably, and harden them against reward hacking and shortcut solutions","Build high‑fidelity evaluation environments with seeded failure modes that expose subtle regressions, and reliable model and agent harnesses with bounded behaviour and explicit output verification","Turn Replay's de‑identified data into realistic state, tools, files and context for an agent, and keep environments reproducible, observable and safe to roll out, with clear evidence of what changed between versions","Work directly with frontier lab researchers to understand capability gaps and design task suites that expose and train against them","Measure task difficulty and solvability by running frontier models against environments, then tune them until they give a usa
CoT_job_summary_short":"COMPANY NAME: Replay\nKEY POINTS: Production RL environments, real enterprise data, ownership mindset"
job_summary_short":"<p>Replay is seeking an RL Environment Engineer to turn real enterprise data into production‑grade training environments for frontier AI labs. You will design tasks, tools, and evaluators, partnering with researchers to close capability gaps and ensure reproducible, safe rollout of environments.</p><p>The role demands shipping ownership, strong software engineering, and experience building environments used for real training/evaluation, with a focus on robustness and measurable signal quality.</p>
contract_type(
fulltime
location_type
on site
remote_scope
null
perks
[{"name":"Meaningful equity"},{"name":"Medical, dental & vision"},{"name":"Unlimited PTO"},{"name":"Gym memberships"},{"name":"Desk allowance"},{"name":"Relocation bonus"}]
stated_min
220000
stated_max
320000
stated_absolute
null
salary_estimation_cot
RL Environment Engineer in New York: base salary around 220k–320k USD with meaningful equity.
estimated_low
220000
estimated_medium
270000
estimated_high
320000
currency
USD
frequency
yearly
bonus
null
tax
gross
application_email
null
job_end_date
null
contact_person
null
hiring_department_name
null

Responsibilities

Design and build production‑grade RL environments grounded in real enterprise workflows, from task spec through to a stable, versioned release
Write reward functions and programmatic verifiers that score outcomes reliably, and harden them against reward hacking and shortcut solutions
Build high‑fidelity evaluation environments with seeded failure modes that expose subtle regressions, and reliable model and agent harnesses with bounded behaviour and explicit output verification
Turn Replay's de‑identified data into realistic state, tools, files and context for an agent, and keep environments reproducible, observable and safe to roll out, with clear evidence of what changed between versions
Work directly with frontier lab researchers to understand the capability gaps they want to close, and design task suites that expose and train against them
Measure task difficulty and solvability by running frontier models against environments, then tune them until they give a usable training signal
Ship quickly, and help set the standards and patterns for how Replay builds tasks and environments

Skills

ML / software engineering

Production‑grade RL environments

Turn researchers' goals into systems

Environment design tradeoffs

Evaluation harnesses / rigor

Ownership mindset

Startup experience

Communication with researchers & peers

Tools

RL environments tooling

Job description

About Replay

At its core, Replay was founded to help founders. We started by supporting startups through shutting down, but we've since expanded into unlocking a new revenue stream for all types of businesses.

In 2025, we earned a secret: the data every company generates on a daily basis (collaboration, communication, building) is some of the most valuable training data in the world. Public and synthetic can only get frontier models so far, so the next generation of model progression depends on real, proprietary data grounded in how actual businesses operate. We're a primary source of it, partnering directly with the frontier AI labs building what's next.

Why join Replay now:

We've scaled from $0 to a multi-eight figures run rate in a matter of months
We've raised from top-tier investors (Floodgate, Afore, Ludlow, Hustle Fund, etc.)
We're small enough that you'll carry outsized responsibility and grow as fast as the company does
You'll partner and build for the fastest and most important companies in the world
You'll be a part of building a massive, category-defining business from the ground floor

The Role

Replay is hiring an RL Environment Engineer who has already built production-grade RL
environments and wants to point that craft at real enterprise data to shape our datasets into the training environments frontier AI labs use to make their models better. You'll turn real business workflows and real-world data into realistic, verifiable environments, with the tasks, tools, and graders to match, and work directly with lab researchers to understand which capabilities they want to improve. This is a hands‑on engineering role with a high degree of ownership.

What You'll Do

Design and build production-grade RL environments grounded in real enterprise workflows, from task spec through to a stable, versioned release
Write reward functions and programmatic verifiers that score outcomes reliably, and harden them against reward hacking and shortcut solutions
Build high‑fidelity evaluation environments with seeded failure modes that expose subtle regressions, and reliable model and agent harnesses with bounded behaviour and explicit output verification
Turn Replay's de‑identified data into realistic state, tools, files and context for an agent, and keep environments reproducible, observable and safe to roll out, with clear evidence of what changed between versions
Work directly with frontier lab researchers to understand the capability gaps they want to close, and design task suites that expose and train against them
Measure task difficulty and solvability by running frontier models against environments, then tune them until they give a usable training signal
Ship quickly, and help set the standards and patterns for how Replay builds tasks and environments

Required aptitude

3-5+ years of ML / software engineering experience shipping production systems
Personally built production‑grade RL environments that were used in real training or evaluation, not only prototypes or research demos
Ability to turn ambiguous goals from researchers into well‑scoped, working systems
A clear grasp of environment design tradeoffs: task specification, state and tool design, determinism, difficulty calibration and solvability
Built evaluation or agentic harnesses, with comfort in evaluation rigor: precision, recall, calibration, imperfect labels and distribution shift
A high‑ownership, action‑oriented mindset: you take problems from idea to shipped.
Startup experience, a taste for broad ownership and the ability to do well when requirements are evolving
Clear written and verbal communication with researchers, engineers, and non‑technical partners

Bonus

Background at or working with a frontier AI lab, or in research‑adjacent engineering
Experience in forward‑deployed or customer‑facing engineering roles
Experience with sensitive enterprise data or other high‑trust production systems
Previously a founder or early employee at a startup

Compensation & Benefits