Founding Research Engineer

Ambral (YC S25); United States; On-site; USD 215,000 - 330,000; Full time

Benefits offered by this job

Significant equity

Equinox membership

Free meals, coffee, and snacks

Health insurance

Unlimited PTO

Job summary

Ambral is building a replayable environment engine over real enterprise history. The system reconstructs a company’s context from past data, exposing it through production-like tools so new policies and agent configurations can be tested against real outcomes.

You will own the research and infrastructure to turn this into a scalable model-improvement system, tackling environment design, graders, and evaluation while collaborating closely with the CTO on deployments.

Qualifications

• 4+ years of experience building production software or machine-learning systems.
• 2+ years working on reinforcement-learning environments, LLM post-training, evaluation infrastructure, or related systems.
• Ability to translate fuzzy business objectives into measurable tasks and signals.
• Ability to diagnose model limitations and separate causes among model, context, tools, harness, or training.
• Capability to move between research questions and production implementations.

Responsibilities

• Build an environment factory turning recorded data and task definitions into runnable environments.
• Design graders to turn business objectives into verifiable rewards.
• Mine useful tasks, trajectories, and evaluation cases from historical workflows.
• Create eval sets that are representative, reproducible, and resistant to overfitting.
• Find model/tool/context/policy combinations to maximize performance while reducing inference cost.
• Train and evaluate agents operating over long horizons and large tool spaces.
• Build replay and observability systems for explainable, measurable agent behavior.
• Scale from thousands of concurrent training and evaluation runs.

Skills

RL environments

Evaluation infra

Agent harness

Production software

Job description

What we do

Ambral helps enterprises own the intelligence behind their most important workflows.

Every company has years of historical evidence showing how work gets done: the context people had, the decisions they made, the actions they took, and the outcomes that followed. Today, most of that history is inert. It isn’t structured in a way that companies can use to evaluate models and improve agent behavior.

Ambral turns this history into replayable environments and eval sets grounded in real workflows and observed outcomes. We use those environments to improve model performance through reinforcement learning and other post-training techniques, alongside context engineering, harness design, and agent engineering.

The result is better, more cost-efficient AI for each enterprise’s specific work, powered by open-weight models that the company owns and controls. This allows each company to retain ownership of its core workflow intelligence instead of outsourcing it to a model provider.

We graduated from Y Combinator Summer 2025, raised millions in funding, and are already deploying within multi-billion-dollar enterprises. Now we’re growing the founding team.

What you’ll do

We’re building a replayable environment engine over real enterprise history.

The system reconstructs a company’s context as it existed at any past time, then exposes that state through the same tools an agent would use in production. This lets us place new policies and agent configurations inside real historical environments, observe how they reason and act, and grade their performance against real outcomes.

You’ll own the research and infrastructure required to turn this into a scalable model-improvement system. The core problems include:

• Building an environment factory that converts recorded enterprise data and task definitions into runnable environments
• Designing graders that turn ambiguous business objectives into verifiable rewards
• Developing methods for mining useful tasks, trajectories, and evaluation cases from historical workflows
• Creating eval sets that are representative, reproducible, and resistant to overfitting
• Finding the right combinations of models, tools, context, and policies to maximize performance while reducing inference cost
• Training and evaluating agents that operate over long horizons, incomplete information, and large tool spaces
• Building replay and observability systems that make agent behavior explainable and measurable
• Scaling from individual environments to thousands of concurrent training and evaluation runs

These problems are wide open. You’ll have significant ownership over both the research direction and the production systems that make it real.

You’ll work directly with the CTO, deploy into real enterprise workflows, and see your research tested against consequential problems and observable outcomes.

Who you are

• You have 4+ years of experience building production software or machine-learning systems, including at least 2 years working on reinforcement-learning environments, LLM post-training, evaluation infrastructure, agent harnesses, or closely related systems
• You understand how environment design, reward design, context, tooling, and policy behavior interact
• You’re comfortable turning fuzzy business objectives into tasks and signals that can be evaluated reliably
• You can diagnose whether a model’s limitations come from the model itself, its context, its tools, its harness, or its training
• You can move between research questions and production implementation without treating them as separate jobs
• You write strong software and can build systems that process large, messy datasets at scale
• You care about reproducibility, observability, and understanding why a model behaves the way it does
• You’re looking to do the best work of your life and build something you’ll be proud of for decades

We care much more about what you’ve built and how you think than credentials or conventional career paths.

Benefits

• Significant equity and ownership
• Equinox membership
• Free meals, coffee, and snacks
• Health insurance
• Unlimited PTO

Compensation Range: $215K - $330K