claude codeLocation: Beijing (Haidian) / Silicon Valley. Open to candidates relocating to / building their careers in China.
Compensation: RMB 500K–700K for new grads; higher for exceptional candidates.
JD A —【 Forward Deployed Engineer, AD Development】
One line: You embed with our autonomous-driving algorithm engineers and make them faster — not by auditing them, but by building tools alongside them and bringing AI into their daily engineering loop.
Why this role exists
We build autonomous driving. That means dozens of distinct engineering problems — perception, planning, data pipelines, simulation, labeling, model training. Our algorithm engineers are world-class at the domain. This role brings the other half: AI-native engineering fluency, embedded right next to them.
What you'll do
- Embed with AD algorithm teams across different problem areas; learn their real development loop firsthand.
- Build internal tools and agents that remove friction from their workflow — eval harnesses, automation, custom dev tooling.
- Bring modern AI coding tools (Cursor, Claude Code, agents) into the team's daily practice, and get engineers to actually adopt them.
- Co-design the team's LLM/agent stack and dev process — you're in the room for the decisions on which components and workflows we adopt.
- Turn what works into reusable tools and playbooks other AD teams can pick up.
You must have
- A strong CS / software-engineering foundation from a top US program; ≤3 years out of school.
- A real engineering bar — you direct AI, you're not replaced by it. Your code is better than what a model produces unguided.
- Deeper fluency with LLMs and agents than most senior engineers — with artifacts to show (personal projects, AI tools, agents, automations, or content).
- Comfort with ambiguity and unfamiliar, messy technical domains. Fast learner.
Nice to have
- Exposure to AD / robotics / ML systems.
- Open-source work or a public presence in AI communities.
The setup
You won't be dropped in alone. You're paired with a senior domain expert who brings the engineering depth and domain credibility; you bring the AI fluency. Bidirectional — you each level the other up.
JD B —【 Forward Deployed Engineer, Business】
One line: You root into one pilot business unit, learn it cold, and ship production software that moves a real business metric — with full authority to write production code, not just demos.
Why this role exists
Our business platforms (e.g. our RAS platform and production systems) are where AI either creates real leverage or stays a slide deck. This role goes deep into one unit and owns the outcome end-to-end.
What you'll do
- Embed deep in one business unit. Before you write a line of code, you go deep on the business.
- Own end-to-end delivery: write production code, modify the engineering repos, ship to production. You hold real commit authority here — this is not a demo role.
- Build LLM/agent-powered solutions that move a metric the business actually cares about.
- Iterate fast on frontline feedback; own the result, not the recommendation.
- Codify what works into repeatable patterns and feed the field back into product and engineering.
You must have
- A strong CS / software-engineering foundation from a top US program; ≤3 years out of school.
- The ability to ship production software solo, in an ambiguous environment — something that didn't exist before you, now used by real people.
- Deeper fluency with LLMs and agents than most senior engineers — with artifacts to show.
- Strong business empathy: you can step into an unfamiliar business scenario and quickly think from the end user's side.
Nice to have
- Founder / 0-to-1 / startup experience, or a product/ops internship with an end-to-end launch you owned.
- A track record of taking ideas to working, shipped products.
The setup
You're paired with a senior domain expert in the business unit who gives you the domain judgment and the internal credibility to move fast without breaking trust. You bring the AI-native build speed.