Back to featured work

    Case study

    AI Agent for Legal Research & Submission Workflows

    Citation-first AI agent for legal research, argument strategy, and submission drafting over authoritative public sources.

    MVP

    The problem

    Lawyers needed to research legislation and court rulings, develop argument strategies, and draft submissions with sources they could verify. Generic chat tools were not enough: answers had to cite authoritative documents and support professional review.

    What I built

    • RAG pipeline over public legislation and court decision corpora
    • Tool-using agent workflows for retrieval, citation, and structured outputs
    • IDE-like research UX: chat with the agent and open cited PDFs from the conversation
    • Submission and argument-strategy assistance grounded in retrieved sources
    • Iterative retrieval and prompt design toward auditable legal research assistance

    My role

    • Solo builder: architecture, agent orchestration, RAG, and application UI
    • Co-developed with practicing lawyers through customer discovery on a representative matter
    • Production-minded design: guardrails, structured outputs, and eval-oriented iteration

    Screenshots

    Matter mapping — agent inventories filings, reads key documents, and proposes an attorney task plan
    Matter mapping — agent inventories filings, reads key documents, and proposes an attorney task plan
    Court-decision research subagent — citation-backed supporting points, limits, and analyses in the Research phase
    Court-decision research subagent — citation-backed supporting points, limits, and analyses in the Research phase

    Stack

    PythonTypeScriptReactElectronNode.jsRAGAI AgentsOpenAI APILLMs

    Status & learnings

    Validated that citation quality and source grounding matter more than fluent prose in legal workflows. The MVP focused on research methodology and professional review, co-developed with lawyers rather than shipping a consumer-facing product.