Ryan McLaughlin

Software Engineer · AI Systems

Sous-chef turned software engineer — still all about mise en place.

Full-stack engineer with 5 years across Java, TypeScript, and Python — now focused on multi-agent orchestration with LangGraph, custom MCP servers, and human-in-the-loop workflows backed by real observability.

01

About

I build production AI systems — multi-agent orchestration with LangGraph, custom MCP servers, retrieval, and the tracing and human-in-the-loop controls that keep them reliable. I'm a software engineer at Paychex working on internal enterprise AI infrastructure, with five years of full-stack engineering (Java, TypeScript, Python) underneath it.

I got here after ~15 years in management and recruiting and a BS in Management Information Systems from Iowa State. I ship, I move fast without breaking the things that matter, and I go deep enough on the stack to debug it at 2 a.m.

Off the clock I cook (professionally, once upon a time), get out on the trails, and generally chase good food — so the meal-planner project below wasn't exactly a coincidence.

02

Skills

AI Engineering

  • LangGraph (multi-agent orchestration)
  • LangChain · LangSmith tracing
  • Model Context Protocol (custom MCP servers)
  • Hybrid retrieval / RAG — vector + keyword search, eval (recall@k, MRR)
  • Multimodal LLM input — vision & in-browser speech
  • Human-in-the-loop & structured LLM output

Languages

  • Java
  • TypeScript / JavaScript
  • Python
  • Go
  • SQL
  • HTML & CSS

Frameworks & Backend

  • Spring Boot
  • FastAPI
  • Angular
  • React
  • Node.js

Data & Infrastructure

  • MongoDB
  • Kafka
  • Kubernetes
  • Docker
  • SQLite / PostgreSQL · sqlite-vec (vector search)

Practices

  • Agile / CI-CD
  • Test-driven development
  • Git & trunk-based workflows
  • On-call / production ownership
03

Projects

Flagship · Open source

Mise-en-Place

A multi-agent meal-planning system for athletes — built to show what production-grade agent orchestration actually looks like.

A LangGraph graph coordinates specialized agents to plan a week of macro-targeted meals, pausing for a human-in-the-loop review before anything is saved. A hybrid retrieval layer (vector + keyword search, fused and measured with a recall@k / MRR eval) powers a cited Nutrition Coach and grounds the plans themselves in real sports-nutrition research and recipes. Add pantry items by typing, snapping a photo (LLM vision), or speaking — Whisper transcribes in the browser, no key and no upload. Custom MCP servers, parallel per-day generation, and LangSmith tracing throughout; bring-your-own-key for Claude or OpenAI.

  • LangGraph
  • RAG · sqlite-vec
  • MCP
  • Vision + Voice
  • FastAPI
  • React + TS
  • LangSmith

Live demo is bring-your-own-key: connect a Claude or OpenAI key to generate plans. Hosted free — first load may take ~30s to wake.

Architecture highlights

  • Hybrid RAG — sqlite-vec (dense) + FTS5 BM25, fused with Reciprocal Rank Fusion — measured by a recall@k / MRR eval harness
  • One retrieval engine powers the cited Coach, plan grounding, and recipe selection
  • Multimodal pantry: photo via LLM vision, voice via Whisper running fully in-browser (free, key-free)
  • Multi-agent LangGraph orchestration with a checkpointed graph + human-in-the-loop interrupt
  • Custom MCP server for pantry & grocery state; parallel per-day generation
  • Local embeddings & in-browser transcription — retrieval and voice need no API key
  • 139 backend tests + an eval regression guard
Side project · Private

webex-assistant

An always-on AI "chief of staff" for Webex, written in Go. It watches incoming messages and triages each one with an LLM — ignore, auto-reply, alert, or defer — clearing the noise and surfacing only what actually needs me. Human-in-the-loop approval before anything sends, LLM tool-calling to search and send messages, a feedback loop that learns what to ignore, and REST + WebSocket APIs driving a live frontend. Private — happy to walk through the architecture.

  • Go
  • Agentic LLM
  • Tool-calling
  • WebSocket
  • SQLite
  • Webex API
Professional · Private

Service Concierge

Internal enterprise AI platform at Paychex connecting enterprise data sources to LLM pipelines through custom MCP servers and multi-agent orchestration, with LangSmith observability and human-in-the-loop checkpoints. Proprietary — happy to discuss the architecture in conversation.

  • LangGraph
  • LangChain
  • MCP
  • Enterprise
Earlier work

Foundations

A trail of pre-AI builds — a Java loyalty/POS back-end, a Java/Swing time-clock app, and a Python + SQL retail price finder. Useful for tracing how I got here.

04

Experience

  1. 2022 — Present

    Software Engineer · Paychex

    Build production AI infrastructure — the "Service Concierge" platform — connecting enterprise data to LLM pipelines via custom MCP servers and multi-agent LangGraph orchestration, with LangSmith observability and human-in-the-loop design.

  2. 2022

    Software Developer · Centene (contract)

    Intranet web services and back-end database support for internal business partners — Angular/TypeScript front ends over MongoDB and SharePoint, Karma/Protractor testing, Node.js, Agile CI/CD.

  3. 2021 — 2022

    Software Developer · Berkley Technology Services

    Full-stack engineer on a billing-services platform supporting 50+ divisions — Java back end, Angular/HTML/CSS front end, rotating 24-hour on-call production support.

  4. 2021

    Software Developer · The Palmer Group (contract)

    Front-end development for a smart-home-services client — TypeScript/Angular, paired programming and TDD, MVC, Agile.

  5. Prior

    ~15 years in management, recruiting & operations

    Retail and store management, IT/financial-services recruiting, and business-transition consulting before pivoting into software — where the soft skills still pay off daily.

05

Contact

I'm open to senior AI and software engineering roles. The fastest way to reach me is email.