
Staff Software Engineer, Agents
Harvey
Posted 2026-05-01
USD 231,000 - USD 340,000 per year
Tech & Engg
Job Description
Ovii's Interpretation of the Role
As a Staff Software Engineer on the Agents team, you will create AI‑driven legal assistants, design evaluation frameworks, and build low‑latency infrastructure that powers high‑impact products for lawyers.
Role Snapshot
- Build AI agents for legal workflows
- Design practical evaluation harnesses
- Optimize model performance and prompts
- Develop low‑latency infra and caching
- Improve observability and instrumentation
- Collaborate with customers and product managers
Must-Have Requirements
- Python
- LLM APIs
- Agent frameworks
- Prompt engineering
- Model selection
- Evaluation design
- Shipping user‑facing products
- software engineering
- LLM / AI agent development
- backend or full‑stack product delivery
Nice-to-Have Signals
- Low‑latency infrastructure (caching, parallel calls)
- Observability & instrumentation
- Passion for domain‑specific agents
- Iterative mindset
Work Setup
- Location: San Francisco, USA
- Work mode: HYBRID
- Remote scope: UNSPECIFIED
- Employment type: Full-Time
Not Specified in JD
- Visa sponsorship
- Relocation
- Travel
- Security clearance
- Coding test
- Portfolio
- GitHub
- Writing sample
- Cover letter
What You'll Likely Work On
- Partner with customers and PMs to map legal workflows and define what “excellent” performance looks like
- Create and iterate on evaluation harnesses that drive agent quality and customer delight
- Apply prompt engineering, model selection, tool design, and context‑window management to boost agent performance
- Design and implement low‑latency execution infrastructure, including caching, parallel tool calls, and sub‑agent patterns
- Enhance observability and profiling to surface bottlenecks and guide optimization decisions
- Stay abreast of the latest research in agentic systems and translate insights into product improvements
Good Fit If You Have
- Passion for domain‑specific AI agents
- Iterative mindset – ship v0s quickly and iterate
- Comfortable using evaluations to drive quality
- Adaptable and open to adopting new best practices
Skills
- Python
- LLM APIs
- Agent frameworks
- Prompt engineering
- Model selection
- Evaluation design
- Low‑latency infrastructure
- Observability & instrumentation
- Backend / full‑stack product delivery