> Markdown version of [/jobs/ext/4620717-founding-ai-engineer](https://www.wearedevelopers.com/jobs/ext/4620717-founding-ai-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Founding AI Engineer

- **Company:** Bronco AI Inc.
- **Location:** San Francisco, CA, United States
- **Contract:** Permanent contract
- **Skills:** Training Data, Artificial Intelligence, Software Bug Management, Computer Engineering, Electronic Design Automation, Python (Programming Language), Reinforcement Learning, Pytorch, Large Language Models, Agentic-AI, Build Tools
- **Published:** November 7, 2026
- **Apply:** https://www.juju.com/job/28_25197_ecb6ec2ab91c92a0c127e6645f9e2000

## About the Role

* Experience at an industry research lab, as a past founder, or as a founding engineer at an AI startup
 * Deep hands-on experience with post-training: RLHF, DPO, or equivalent
 * Experience designing evaluation frameworks for LLM or agentic systems
 * Strong proficiency in Python and PyTorch, * Background in Electrical or Computer Engineering
 * Familiarity with chip design and verification workflows
 * Experience with EDA tools (Synopsys, Cadence, Questa, etc.)
 * Publications at NeurIPS, ICLR, ICML, DAC, ICCAD, or DVCon

## Description

As a Founding AI Engineer, you will own the core intelligence of Bronco's agents. That means designing the training pipelines, reward models, and evaluation harnesses that will make sure our agents actually succeed not just in benchmarks but on real production tapes., * Design and implement post-training pipelines for domain-specific model behavior across DV tasks - testbench generation, bug triage, coverage closure

 * Develop reinforcement learning setups, including reward modeling and RLHF/RLAIF pipelines, to improve agent decision-making on long-horizon verification tasks
 * Build tool interfaces, scaffolding, and execution harnesses that allow models to interact reliably with EDA environments and RTL artifacts
 * Iterate with silicon domain experts and customers to encode verification knowledge into training data, reward signals, and evals that measure real agent capability - not proxy metrics
 * Stay current with frontier research in LLM reasoning, RL for agents, and tool use - and bring what's relevant into production