I'm James Blanchard — I build AI tools that make it into production, not demos. I prototype fast, harden what earns its place, stand up the data pipelines and infrastructure behind it, and stay close to the people who actually use it.
I'm an engineer who likes owning the whole lifecycle — designing the architecture, writing the code, wiring up the database, building the pipeline, and supporting it once it's live. That last part matters to me: I stay on the hook for the systems I ship and the people using them.
Most of my recent work is applied AI — building agent pipelines on the Anthropic API, testing models and frameworks that don't have a proven track record yet, and putting the ones that hold up into production behind real guardrails. I write Python, JavaScript/TypeScript, and Dart among others, run everything through Git and automated CI/CD, and deploy on Cloudflare Workers, D1, and R2.
I've shipped real products solo: a live group food-decision app, a published Play Store app, and an AI agent harness that runs multi-stage builds end to end. Alongside that I handle the unglamorous half — data pipelines, ERP integrations, and day-to-day IT and end-user support — which is usually where you find out what a business actually needs. I learn fastest by building, I teach myself tools before anyone asks, and I document what I ship so other people can run it.
Enterprise IT support for clinical and administrative staff across HCA's national hospital network. I work tickets in ServiceNow from routine incident intake through critical, high-priority escalations — account and access administration in Active Directory, device imaging and deployment, Citrix / VDI, VPN and virtual-desktop troubleshooting, and support for the clinical applications staff depend on to do patient care. Advanced issues get escalated to the right specialized team or vendor with the context already gathered. It's the job that taught me to explain technical work to people who don't want a technical answer, and to stay calm when the system someone needs is the one that's down.
A group food-decision app that helps people land on where to eat together. Flutter/Dart frontend backed by a microservice architecture of Cloudflare Workers, a D1 (SQL) database, and Durable Objects for real-time session state — integrating the Google Places and Anthropic APIs, deployed through CI/CD. One project that spans OOP, relational data, cloud, microservices, version control, and AI tooling end to end.
A managed social-publishing platform that drafts posts in a brand's voice, holds them in an approval queue, and publishes across X, Instagram, Facebook, Threads, Bluesky, LinkedIn, and TikTok. Built human-approved by default — drafts wait for a one-tap sign-off from the phone, with per-brand auto-publish available once the voice is trusted and automated safety checks on every post either way. Multi-tenant by design, with versioned brand-voice prompts, scheduled cron publishing, per-post cost and reach tracking, retry on platform failures, and a full approval audit log.
An agent harness that turns a spec into shipped, verified work — the infrastructure layer for running AI in production. A written PRD enters an approval queue and moves through a multi-stage pipeline — grounding, planning, checkpoint, build, and an independent cross-model judge — with every stage required to produce receipts: the actual command, its output, and its exit code. Claims without them don't pass. An Obsidian vault of markdown is the single source of truth for specs, doctrine, and the decision log, so state survives across sessions and parallel agents don't drift. Builds run in a sandbox against a local SQLite store, dependencies are pinned deliberately rather than drifting, and the human keeps the final gate on anything that ships, spends, or sends — with hard cost caps and a full attribution trail behind it.
A tip-calculator mobile app built in object-oriented Dart and published to the Google Play Store. I owned the full lifecycle — code, store listing, graphics, and beta testing — through public release.
A two-sided marketplace app connecting boat owners with boat-cleaning services. Working alongside the lead developer, I built frontend features and UI/navigation flows in Dart and JavaScript, and helped implement authentication for Google, Apple, and email sign-in. Contributed through pair programming, code reviews, and sprint participation in an agile team.
A website-build service delivering modern, responsive sites for small businesses end to end on Cloudflare Pages, with Git-based deployment. Covers the full build from design through shipping and hosting.
I'm open to AI and software engineering roles and collaborations. If you're hiring or have a project in mind, I'd love to hear from you.