Research Engineer

Oho Group; San Francisco (CA); On-site; USD 160,000 - 230,000; Full time

Job summary

Oho Group is seeking an AI Researcher to advance post-training methods and agentic RL for complex hardware-engineering workflows. The role emphasizes hands-on development and practical deployment of AI agents in real-world tasks.

You will implement RL environments, generate synthetic data, and build benchmarks while collaborating with hardware experts to translate research into production-ready tools.

Qualifications

• Proven experience with LLM post-training and RL environments.
• Evidence of both research and practical implementation of AI agents.
• Experience using Cursor or Claude Code in development workflows.

Responsibilities

• Develop post-training approaches for agents performing multi-step tasks.
• Build RL environments with executable tasks and reliable reward signals.
• Generate and filter synthetic data using agent feedback to guide training.
• Design benchmarks testing tool use and generalisation.
• Collaborate with hardware specialists to map workflows to learnable tasks.
• Turn research into working code and contribute tools/infrastructure.

Skills

Software engineering fundamentals

Post-training ML / RL

AI agents

Tooling (Cursor / Claude Code)

Tools

Cursor

Claude Code

Job description

AI Researcher — Post-Training & Agentic RL

Join a well-funded startup developing AI agents for complex hardware-engineering workflows. The team brings experience across advanced AI research and semiconductor engineering, with work connected directly to frontier-model development.

They’re looking for a hands‑on AI Researcher with strong post‑training experience to help agents reason, use tools and complete increasingly difficult technical tasks.

There is substantial room to shape the research: deciding which tasks agents should learn, building environments that provide useful feedback, and turning experimental results into systems that work on real engineering problems.

Your work would include

• Developing and evaluating post‑training approaches for agents working through multi‑step engineering tasks.
• Building RL environments with executable tasks, reliable reward signals and meaningful measures of success.
• Generating, selecting and filtering synthetic training data, using agent failures to identify what the next training cycle needs.
• Designing benchmarks that test correctness, tool use and generalisation to unfamiliar problems.
• Working with hardware specialists to translate engineering workflows into tasks agents can learn from.
• Taking research into working code, contributing to the supporting tools and infrastructure where needed.

About you

• You’ll need strong software engineering fundamentals, practical experience with LLM post‑training and RL environments, and evidence of both research and implementation involving AI agents.
• You should also be comfortable using tools such as Cursor or Claude Code in your own development workflow.
• Experience with digital design, VLSI, FPGA, EDA or verification would be a significant advantage. Academic projects or a serious personal interest in hardware are also relevant.