Staffing for AI labs and data vendors

Hire the engineers who build RL environments

Task authors, verifier engineers, environment infrastructure and QA and calibration. Production engineers first, environment builders second, under one EltexSoft contract.

Request the roles you need What the pod builds

EltexSoft staffs the four roles an RL environment needs: task authors who mine real codebases into SWE tasks, verifier engineers who write the hidden tests and reward rules, environment infrastructure engineers who build and isolate the Docker images, and QA and calibration engineers who try to break every task before a model sees it. Contract the whole pod under one agreement or add single roles to your own environment team. Rates on request.

Roles

Four roles, one environment

An environment is only as good as its weakest role. Each one below is a distinct job, and the person who writes a task never grades it.

Engagement models

Three ways to contract them

The pod, under one contract

The pod covers all four roles under one contract, with EltexSoft running the queue. You set the task profile and the format, you grade the output. Fixed-price pilots.

How the pod delivers

Single roles on your environment team

One or more of the roles above under your own tech lead, in your repository and your tooling. Staff augmentation terms: you direct the work, we handle employment, retention and replacement.

Staff augmentation terms

A dedicated environment team

Your own pod with an EltexSoft lead, working your backlog of environments full time. The right shape when you have more codebases than one pod can turn into tasks.

Dedicated team terms

Before day one

What every engineer clears

  • Identity and location verified before any repository access
  • NDA and IP assignment in place before the first task
  • Project-only access, synthetic data, no production systems
  • Professional liability insurance, certificate on request

Rates

Quoted per role, on request

Tell us the roles, the format (Harbor, HUD or both) and whether you want finished tasks or hours. You get a per-role quote and, for pod work, a fixed pilot price before anything starts.

Ask for rates

FAQ

Common questions

What does an RL environment engineer do?
An RL environment engineer turns a real codebase into training material for coding agents. Depending on the role, that means writing tasks and gold solutions, designing the hidden tests that compute the reward, building the isolated Docker image the agent runs in, or validating and calibrating tasks before delivery. The common thread is production software experience: the tasks are only as realistic as the engineer who has shipped that kind of code.
Which role do I need first?
If you already have an environment and need more tasks, start with a task author and a verifier engineer working as a pair; one writes, the other grades. If you are starting from a bare codebase, the environment infrastructure role comes first, because nothing can be validated until the image runs deterministically. QA and calibration is the role most teams add last.
Can I hire one engineer rather than the whole pod?
Yes. Single roles are available on staff augmentation terms: the engineer works under your tech lead, in your repository, with your tooling, and you direct the work. The pod under one contract is the better fit when you want finished tasks rather than hours, because the four roles check each other and the validator and calibration steps are already in place.
How fast can an RL engineer start?
The start date is agreed per role when we scope the engagement. Pod engagements start with a fixed-price pilot on an environment that already exists, so the first tasks arrive within the pilot window rather than after a ramp-up. Single roles on your own environment need onboarding into your repository first, and that time is part of the plan, not a surprise.
What are your rates?
Rates are quoted on request, per role and per engagement model, once we know the task profile and the format you need. Pilots are fixed price so the first engagement has no open-ended cost.

Tell us which roles you are missing.

An engineer replies within one business day with availability and a per-role quote.

Request the roles