Emilio Esposito

    Senior Director, AI Engineering at LegalZoom
    Co-founder & Managing Partner, Sernia Capital

    Emilio Esposito

    I lead AI Engineering at LegalZoom, where I built the function from the ground up and still write production code daily: shipping agentic systems, deploying LLM-powered products, and driving AI tooling adoption (including Claude Code) across a 200-person engineering org.

    I also co-founded and operate Sernia Capital, a 40-unit residential real estate portfolio. This site serves double duty: it hosts production services that power Sernia Capital's day-to-day operations, and it's a sandbox where I experiment with new technologies in my free time.

    ResumeGitHubLinkedIn

    Sernia Capital: AI-Powered Operations

    Every system below was designed, built, and deployed by me to solve real operational problems across our portfolio. These run in production 24/7, handling everything from emergency escalation to lead qualification to tenant communications.

    Sernia AI: Operations Assistant

    Production

    Multi-modal AI assistant that operates across SMS, email, and our web app. It continuously reads all business communications, autonomously learns over time through a memory-based feedback loop, and can take actions (sending messages, managing tasks, scheduling jobs) with human-in-the-loop approval. Triggered both by incoming messages and on a scheduler so it proactively follows up with property managers on project status. Built with PydanticAI.

    Emergency SMS Routing

    Production

    Agentic AI monitors every inbound tenant message in real-time. Urgent issues trigger automated call/text escalations via Twilio that bypass Do Not Disturb during off-hours.

    View source

    Intelligent Lead Management

    Production

    AI monitors Zillow email threads, extracting contacts, creating calendar events, syncing to our phone platform, and triggering follow-up reminders for leasing agents.

    View source

    Tenant Communications

    Production

    Bulk SMS messaging for building-wide announcements, built to work around platform limitations with role-based access control.

    Open app

    AI Leasing Auto-Replies

    In Development

    Agentic AI auto-responds to inbound leasing inquiries with full context on properties, listings, and agent availability. Screens applicants before proposing meeting times.

    Preview

    Favorite Stack

    Tools I reach for first when starting something new.

    AI & Agents

    PydanticAI

    The right level of abstraction for building agents: structured outputs, dependency injection, and tool calling without fighting the framework. The Graph Beta API is excellent for multi-agent workflows.

    FastMCP

    The FastAPI of MCP servers. Makes it trivial to expose tools, resources, and prompts over the Model Context Protocol. I've built MCP servers for LegalZoom and use them heavily in my own workflows.

    Dev Environment

    Claude Code

    My go-to for tackling larger features end-to-end. Changed how I write software. I use it for everything from greenfield projects to navigating massive legacy codebases. Led its rollout across 200 engineers at LegalZoom.

    Cursor

    My daily driver IDE for when I want to be in the driver's seat with a bit of AI assistance (tab autocomplete, inline suggestions) to keep the flow smooth without giving up control.

    Frameworks

    FastAPI

    Type-safe Python with async support and auto-generated docs. Pairs perfectly with Pydantic models and makes building AI backends genuinely enjoyable.

    React Router v7

    Remix rebranded. Loaders, actions, nested routes, and the Vite bundler makes hot-reloading noticeably faster than Next.js.

    Infrastructure

    Railway

    Deploy anything without thinking about infrastructure. Monorepo support, PR preview environments, and a CLI that actually works. Their MCP server makes deploying and debugging from Claude Code seamless.

    Logfire

    Observability from the Pydantic team, built for Python and AI workloads. Native PydanticAI tracing out of the box. Their MCP server lets me query traces and debug production issues without leaving the terminal.

    Neon Postgres

    Serverless Postgres with branching. I spin up isolated database branches for every PR environment. Zero config, zero cleanup.

    Full source for this site is open on GitHub.