The role
- COMP
- $200K - $300K
- EQUITY
- Competitive
- LOCATION
- San Francisco · New York
- WORKPLACE
- Hybrid
- EXPERIENCE
- 4+ years
- VISA
- None, Visa transfers
- STACK
- React, TypeScript, Python, PostgreSQL, GraphQL, FastAPI
- INDUSTRY
- AI, Law, B2B
/roles — ROLE_275
Senior / Staff Software Engineer, AI Systems
Seed-stage immigration company using AI agents to draft filings and move visa cases forward for tech talent
The company
Immigration startup helping people in tech obtain US visas quickly, combining an in-house legal team with AI-powered case software.
- STAGE
- seed-stage
- FUNDING
- seed funding
- TEAM
- ~40 people
- FOUNDED
- 2024
JD — the work
About the role
You would own the AI agents, plus the infrastructure under them, that handle intricate legal work at an immigration startup: preparing documents, assessing evidence, and advancing cases to filing. That covers orchestration of workflows, the backend, and evals that monitor quality and flag regressions. You would work closely with the legal team, learning their processes firsthand and hardening early experiments into production systems. The team is spread across New York and San Francisco, and the company sponsors visas.
What you'll do
- Create agents that carry multi-step legal processes forward, pausing for human checks where they matter
- Rework orchestration so long-running jobs handle dependencies, retries, and failure recovery cleanly
- Stand up evaluation systems that score results, flag regressions, and steer prompt, model, and agent improvements
- Strengthen backend architecture across services, scheduled jobs, and event pipelines
- Add observability to agent runs for debugging and for tracking cost, latency, and quality
- Wire agents into internal tools and the applicant portal, with APIs and screens for human oversight
What they're looking for
- 4+ years of production engineering with strong backend depth and experience building AI or ML applications
- Built with LLMs past simple chat, thinking hard about context management, tool calls, evals, and flaky outputs
- Skill at translating experts' subtle quality judgments into code and useful evaluations
- Fluency in TypeScript and Python or comparable languages
- Strong understanding of async systems and orchestration, including idempotency, queuing, retries, and events
- High agency, having owned projects from unclear requirements through launch and day-to-day operation
Nice to have
- Production agent operations, eval tooling, or document-heavy workflow experience