Sean Newman — AI and software engineering

What I build

Most engagements are a fixed scope with a delivery date attached. Where the problem isn't clear yet, we start with a short paid diagnostic and a written scope before anything larger gets committed to.

Voice AI

Agents that answer, qualify, book and route calls without sounding like a phone tree. Latency budgeted across speech, model and playback, because the delay is what gives them away.

Agentic workflows

Systems that decide and act rather than just respond. Tool use, multi-step reasoning, human checkpoints where the stakes justify one, and traces you can read when something goes wrong.

Automations and integrations

Connecting the systems a business already runs — booking platforms, CRMs, practice and field software, billing. n8n, Make, GoHighLevel or custom, with the auth, retries and failure cases nobody scopes for.

RAG and document intelligence

Retrieval over contracts, records, manuals and archives. Built so the answer cites its source and abstains when the source isn't there, rather than inventing something plausible.

Data pipelines

Collection, enrichment, deduplication and scoring at volume. Resumable and rate-limit aware, so a run that dies at hour six picks up at hour six.

Applied machine learning

Forecasting, classification, ranking and scoring on real business data. Including the unglamorous part — establishing whether the model actually beats the simple baseline.

The product around it

Dashboards, portals, internal tools and the deployed application. AI features are rarely the whole job; something has to hold them, and someone has to be able to operate it.

Something else entirely

If it's software and it's hard, it's worth a conversation. I'll tell you if it isn't a fit for me.

How I work

These hold on every engagement, whether you're the business or the agency in front of it.

Dates You get a delivery date and I hold to it. If something is going to slip, you hear it early rather than on the deadline.
Communication Same business day, usually within a few hours. Written updates you can forward without translating them first.
Scope Fixed and written down before work starts. Changes get quoted separately, in advance. No surprise invoices.
Ownership Work-for-hire. The code, credentials and documentation are yours on completion, and I hand over a system someone else could maintain.
Pace One developer covering ground that usually takes a team. My own workflow is deeply AI-assisted, which is how that arithmetic works.
Honesty If the thing you're asking for won't solve the problem you have, I'll say so before taking the work rather than after delivering it.

If you're an agency

I work white-label. Your client never learns I exist — my name appears in no deliverable, no commit, no email and no call, and I don't contact or market to them afterwards. I sign your NDA before scoping rather than after.

I write the code myself; nothing is subcontracted onward, so you're never more than one person away from whoever is actually building the thing. Standing monthly capacity is available for agencies sending work regularly.

Background

A decade of production engineering and a master's in AI. Research applications at a university before this, including reinforcement learning in healthcare settings — mostly useful now as a habit of not trusting a system until it has been measured.

Languages — Python, TypeScript, JavaScript, Node.js, SQL, Bash
Web — React, Next.js, FastAPI, Express, REST and webhook APIs, WebSockets
AI — Claude and OpenAI APIs, tool use and MCP, agent frameworks, prompt and context engineering, evaluation harnesses
Retrieval — embeddings, chunking strategies, reranking, pgvector and vector stores, OCR and document parsing
Voice — Retell, Vapi, Twilio, speech-to-text and text-to-speech pipelines, latency budgeting
Machine learning — PyTorch, scikit-learn, pandas, forecasting and classification, reinforcement learning
Automation — n8n, Make, GoHighLevel, Zapier, custom orchestration, scheduled and event-driven jobs
Data — Postgres, SQLite, Redis, ETL pipelines, scraping and enrichment, deduplication and entity resolution
Integrations — Stripe, Twilio and A2P registration, Google Workspace and Places, booking and CRM platforms, OAuth
Infrastructure — Linux and VPS administration, Docker, nginx, Cloudflare, Netlify, queue workers, logging and monitoring
Practice — Git, testing, technical scoping and estimation, documentation and handover

A decade of production work also means time in Java, C, C# and PHP. I read and inherit legacy code without complaint.

Start a conversation

Tell me what you're trying to build and what's in the way. I'll reply the same business day with whether it's a fit and what the first step would look like.

Your name

Company or agency

What this is about

What you're building, and what the timeline looks like