AI & Automation Software Engineer

Jake Morgan

I build AI and automation systems and run them in production.

Five of them are deployed, tested, and documented. The code and the eval numbers are below.

The five pieces

Systems I built, deployed, and tested.

Each system is a live service running on a single VPS behind a hardened tunnel. The primary link on every card points to the GitHub repository, where the README, specs, and evaluation evidence live. Raw endpoints may sit behind authentication, so the repositories are the proof, not the bare URLs.

MICT-FIELD-005

Live

FieldOps

Flagship

Proposal-generation platform (capstone)

My capstone. It turns a project intake into a grounded proposal: schema-checked extraction, retrieval over the cabling corpus, and three agents (Estimator, Writer, and Reviewer) working in a capped review loop. The numbers come from a deterministic calculator, not the model, and nothing ships without human approval.

Deployed and live. The flagship of the portfolio.

TypeScript
Postgres
pgvector
n8n
DeepSeek
Gemma
Cloudflare Tunnel

View on GitHub

jakemorganlabs/field_ops_capstone

MICT-RAG-002

Live

RAG

Grounded document intelligence

Answers questions over a technical cabling-standards corpus and cites the source for every answer. When the documents do not support an answer, it refuses instead of guessing. Ships with a full eval report and a named limitation I did not hide.

Deployed and live. Signed query API.

TypeScript
Postgres
pgvector
HMAC
Evals

View on GitHub

jakemorganlabs/document_intelligence_rag

MICT-AGENT-004

Live

Recon

Multi-agent company research

Researches a company and returns a source-cited dossier. It searches, extracts findings, runs a coverage loop to fill the gaps, synthesizes, and checks every claim against its evidence before it writes anything down.

Deployed and live behind a rate-limited endpoint.

TypeScript
Anthropic
Cassette Evals
Cloudflare Tunnel

View on GitHub

jakemorganlabs/recon_multiagent

MICT-PIPE-001

Live

PIPE

Lead intake and outbound pipeline

Takes inbound leads, classifies them, and routes each one where it belongs: hot leads to Slack and HubSpot, warm leads to a Google Sheet, cold leads to Postgres. Classification runs on Gemma through DeepInfra.

Deployed and live.

Hono
Postgres
n8n
Gemma
DeepInfra
Cloudflare Tunnel

View on GitHub

jakemorganlabs/intake_n_outbound.pipeline

MICT-NODE-003

Published

Shovels

Verified n8n community node

An n8n node that wraps the Shovels US building-permit and contractor API: permit search, contractor search, geo resolution, and pagination handled for you. Published on npm with signed provenance and approved by the n8n Creator Portal.

Published and verified. On npm.

TypeScript
n8n
npm
OIDC Provenance

View on GitHub

jakemorganlabs/shovels_n8n_nodes

About

Self-taught. Deploys like a professional.

I taught myself to build software. No computer science degree. I came up through low-voltage and structured cabling work running Morgan ICT, then moved into AI and automation. That is the work that has produced my strongest systems and the work I want to keep doing.

Everything on this site is a system I designed, built, deployed, and tested myself. I document what works and what does not. Right now I am looking for a junior AI automation engineering role.

Eagle Scout, 2023.

LinkedIn GitHub Email

Contact

If the work fits, talk to me.

Email: jakemorganlabs@gmail.com
GitHub: github.com/jakemorganlabs
LinkedIn: linkedin.com/in/jakemorganlabs