Noah Connolly

Scene 03 · Toolbox

What I build with

Not a badge wall — each of these is carrying production traffic somewhere right now.

AI / LLM

Agentic systems that 30+ managers open every morning.

Claude APIOpenAI APIAgentic pipelinesTool useRAGPrompt engineeringStructured outputEmbeddings

Languages & Frameworks

Six years full-stack — both sides of the wire.

PythonJavaScript / TSC#SQLDjango + ChannelsReactVue / NuxtNode / Express.NETTailwind

Cloud & Infra

CI/CD promotion gates that cut release time 70%.

AWS · EC2 / RDS / S3 / VPCGCPTerraformDockerJenkinsGitHub ActionsRedisPostgreSQL

Integrations & Standards

Bidirectional ERP syncs that survive outages and heal themselves.

WebSocket / ASGIQuickBooksSageNetSuiteAcumaticaHL7 FHIRAuth0 / OAuth2Twilio

Scene 04 · Selected Builds

Problems first, code second

Four builds that follow the same arc: a fuzzy, expensive problem — then the right thing, shipped.

AI · Automation

Autonomous Error Remediation

Every production error waited on a human just to become engineering work.

Fingerprints errors in production and assembles the surrounding context automatically
Generates fix candidates through the Claude API
Opens GitHub PRs — the first human in the loop is the reviewer

Changed what the team considers automatable.

Python · Django · Claude API · GitHub · AWS

AI · Product

The AI Daily Briefing

Operators, supervisors, and CEOs open the same dashboard wanting completely different mornings.

Sat with every role across our customers before writing a line of briefing logic
Two-pass LLM pipeline builds role-shaped scorecards — shifts, trends, remediation recommendations
Delivered on each user's own schedule, tied to the existing permission model

The daily anchor for 30+ managers and schedulers.

Claude API · Django · Redis · WebSockets

Scale · Infra

20,000 Parts a Day

Our biggest plant was 10× anything we'd run — and AWS said "healthy" while operators phoned in outages.

Built the observability that didn't exist: thread state, WebSocket churn, honest health checks
Unwound the failure chain — instance sizing, thread pools, reconnect storms, signal floods, memory fragmentation
Auto-remediation now detects blocked threads and memory pressure and recovers in seconds

Green means healthy now.

AWS EC2 / RDS · Django Channels · Redis · ASGI

Healthcare · Data

Health Record Cloud

A multi-practitioner clinic needed records that speak FHIR — across two clouds.

Architected FHIR data capture and transfer spanning AWS and GCP
Audit logging, document version control, async database management over RESTful APIs
Auth0, Twilio, and Typesense for a secure patient-and-clinician experience

Led and mentored the team that shipped it.

HL7 FHIR · AWS · GCP · Node · NoSQL · Auth0

Scene 05 · Contact

Have the right thing to build?

I'm looking for a team where AI is the product, not the garnish — and where I can own the problem from stakeholder conversation to production. If that sounds like your team, let's talk.

Email me LinkedIn GitHub

The path so far