At Epiq , your work contributes to complex, global legal outcomes. You’ll join a values‑driven community where integrity guides decisions, relentless service sets the bar, and we thrive on big challenges together. We invest in your growth with enterprise‑wide learning and mobility. We celebrate who you are, and we respect life beyond work with flexibility that’s recognized externally. Enabled by modern platforms and AI, you’ll do the most meaningful work of your career and see your impact at scale.
Job Description:
Epiq AI Labs is the innovation and engineering hub behind Epiq’s next-generation AI platform for corporate legal departments and global law firms. Operating with the speed and autonomy of a startup and the resources of a global alternative legal services provider, the team builds intelligent agents, reasoning engines, knowledge systems, and structured workflows for litigation, investigations, compliance, and corporate knowledge work.
We are fast, experimental, and product-obsessed, with high autonomy, high trust, and low bureaucracy. We ship to real legal clients and hold a high bar for correctness, and we use AI tools throughout our own development workflow. The team is deeply technical, highly collaborative, and owns its systems end to end — you design it, you build it, you ship it.
The Opportunity
You will build the agentic systems at the center of Epiq AI Labs’ legal AI platform, including the orchestration runtime used by other teams, the retrieval layer agents reason over, and the evaluation infrastructure used to measure quality. You will own systems from initial design through production operation.
Legal workflows require a high standard of correctness, traceability, confidentiality, and tenant isolation. You will establish the primitives for orchestration, grounded retrieval, and measurable quality that other engineering teams will build on.
Essential Job Responsibilities
Design the agentic platform: orchestration of single agents and parallel sub-agents, durable state and memory across long-running jobs, MCP-based tool integration, and trace-level observability.
Build the safety layer agents run inside: tool registration and permissioning, sandboxing, guardrails and output validation, and defenses against prompt injection from retrieved content.
Build agent workflows with multi-step reasoning, tool use, human-in-the-loop checkpoints, and recovery from partial failure during long-running execution.
Evaluate developments in model capabilities, agent-design patterns, and evaluation practices, and translate them into platform architecture.
Design retrieval-augmented generation systems with hybrid search over legal corpora and citation grounding to specific sources.
Own and evolve the in-house model services behind classification, extraction, and embeddings — including customer-trained custom models with active learning — and integrate them alongside frontier models.
Build model-level and end-to-end evaluation datasets, benchmarks, and regression checks for classification, extraction, retrieval, and agent workflows. Work with legal domain experts to turn real workflows, corrections, and failure cases into quality improvements, with automated evaluation in CI and monitoring of quality, latency, and cost in production.
Establish LLMOps practices for running models reliably in production: model routing and fallbacks, latency and cost budgets, prompt and context versioning, caching, and quality and reliability monitoring.
Use AI coding agents throughout development while owning the outcome — the design, specifications and test plans, code review, and verification of correctness, security, and maintainability — and ship major changes behind automated tests and controlled rollouts, with product and legal-domain validation where appropriate.
Build scalable backend services and APIs, produce technical design documentation, and partner with product, research, and security stakeholders.
Participate in the on-call…