Applied Research Engineer

Last Updated

Full Time

Engineering

San Francisco

Full Time

Engineering

San Francisco

Job Description

The next wave of competitive advantage isn’t better general models. It’s models that understand your business.

At Eragon, we build company-specific AI systems trained on proprietary data, deployed in customer environments, and continuously improving through real-world use. Our models don’t just respond. They learn from interaction and get better over time.

We’ve built a reinforcement learning framework called RLQF (Reinforcement Learning from Query Feedback) that turns real usage into training signal and creates a compounding improvement loop beyond static fine-tuning or RAG.

The Role

As an Applied Research Engineer, you’ll design, train, and deploy models that power real business workflows.

This is not research for its own sake. You’ll work directly with customer data, constraints, and feedback to build systems that perform in production. You’ll own the full lifecycle from problem framing and data design to training, evaluation, and iteration in the wild.

What You’ll Do

  • Train and adapt models: Fine-tune and post-train models on customer-specific data using RLQF and related techniques

  • Close the loop: Turn real user interactions, corrections, and workflows into training signal

  • Own end-to-end systems: Build from data ingestion and curation through training, evaluation, and deployment

  • Evaluate in production: Design evaluation frameworks that reflect real-world performance, not just benchmarks

  • Work with customers: Partner directly with users to understand workflows and translate them into model behavior

  • Ship and iterate: Continuously improve models based on live feedback and measurable outcomes

What We’re Looking For

  • Strong hands-on experience training, fine-tuning, or post-training ML models

  • Experience working with messy, real-world data, not just clean benchmarks

  • Familiarity with reinforcement learning, feedback-driven training such as RLHF or RLAIF, or evaluation systems

  • Ability to move quickly from problem to data to model to iteration

  • Strong engineering instincts and comfort owning systems end-to-end

  • Bias toward shipping and improving systems, not just running experiments

Strong Candidates Also Have

  • Experience fine-tuning or adapting large language models in production

  • Background in agents, tool use, or workflow automation systems

  • Experience working in customer-facing or forward-deployed environments

  • Prior startup or early-stage engineering experience

  • Contributions to open source or side projects demonstrating applied ML depth