* Est. 2008 · Programming since 1976 · New Hampshire
eDataQuest is the practice of James Rascoe — 50 years fitting the pieces together, from IBM iron to AI agents. Real engineer. Real accountant. Fortunate to have had the vantage, still at the keyboard every day, and still taking on work.
* The Arc
Five decades of knowing how the pieces fit together — because I was there when they were first connected.
Watergate and disco on the radio. Business ran on room-sized mainframes. I was fortunate to be in the glass room young — on payroll as a teenager, writing code on IBM iron.
MTV launched, the Wall came down, and computers landed on desks. I picked up C and never put it down.
Seinfeld on TV, AOL in the mailbox. The web went from nothing to everywhere, and I was building on it the whole time.
Reality TV and the first iPhone. Business moved into the browser and onto virtual machines. Got on ESXi early — one of those bets that paid off.
Netflix killed Blockbuster. Company data left the building. I moved clients to the cloud and kept them shipping.
Pandemic, Zoom, ChatGPT. AI showed up in everyone's browser. Same practice, next chapter.
* Five decades · still at it
On payroll writing code in 1976 and still shipping in 2026 — the math works out differently than you might guess. One continuous practice across every platform that came along. Fortunate to have had the vantage, too curious to stop.
* AI Practice
Multi-agent orchestration, document intelligence, and conversational systems running in production — integrated into the business, not bolted on the side. I'm still in charge of the code — AI is a tool, not the engineer. Here's what the practice looks like today.
LangGraphMulti-agent workflows on LangGraph — state machines, tools, and handoffs that actually run in production. Not demo code.
SMS · TwilioAI agents handling real customer SMS over Twilio — intake, triage, and decision-making wired directly into business data.
GeminiIngest arbitrary PDFs and images, extract structured data with Gemini, reconcile against the system of record. Less work, fewer errors.
LangSmithFull-trace visibility with LangSmith — replay every decision, every tool call, every token. Debug AI the way you'd debug code.
Anthropic · OpenAI · GoogleClaude, GPT, Gemini, and local models — picked per task, not per fashion. Cost, latency, and quality are all deliberate choices.
GraphQL · Apollo · PostGraphileAn agent is only as good as the data it can see. Strongly-typed end-to-end — Postgres through PostGraphile, Apollo on the client, GraphQL codegen connecting them. Agents work from a clean, typed business model — not raw tables, not hallucinated shapes.
* Practice Areas
Different entry points — same operating principle: understand the problem, ship what works, cut the ceremony that doesn't serve the outcome.
Builds that actually ship. Full-stack — frontend, backend, the glue that connects them. Picked for what will still be running in five years, not what's trending this quarter.
Kubernetes, ArgoCD, CI/CD, certificate management — the plumbing that makes your system actually yours. Self-hosted or cloud. The point is the lights stay on without a SaaS lock-in you can't escape.
Data pipelines, web analytics, and AI agents wired into the business — not bolted on the side. Patterns earned over decades of fitting data to decisions that had to make money.
Short engagements to get clear on what you're actually building, whether to build it at all, and whether a smaller move gets you the same outcome. No techno-babble, no decks of options.
Part-time technical leadership for founders shipping something real. Help you make money on what you build instead of burning through the raise before it earns. (Back when I started this it was called "rent a CTO.")
* About
I'm James Rascoe. I started very early — hands on a computer since 1968, growing up inside the family's software business. By the time I was a teenager I was on payroll writing code on IBM iron. None of it felt unusual at the time — it was just what I did.
My trade got shaped by the era that raised it. In the 1960s IBM wrote the software that came with the hardware — that's how I learned. After the 1969 unbundling, IBM kept selling into every industry but needed independent shops to write and integrate the software that closed the sale. That's what I've mostly done ever since. The industry breadth anyone sees on the record is actually the opposite of variety — one job that happened to take me into every vertical IBM sold into.
In 1987 I left the family business to run my own. I'd seen networked PCs coming, then the internet coming, and the old model wasn't going to hold. I've been running my own shop ever since — different names, different eras, one continuous practice. "President" and "VP" on the record are my own titles; I've basically always been self-employed.
The less common part of the bio: I'm a real engineer and a real accountant. I read balance sheets and I read assembly. That combination is what keeps me honest about whether the thing being built is going to pay for itself — which is usually the question that actually matters.
I founded eDataQuest in 2008 as the container for the practice. It's not a firm with overhead — it's a partner you bring in when you need the thing actually built. When a job needs more hands, I pull them in from a network of collaborators built up over five decades. AI became one of those collaborators recently; having built production AI agents for years, I know how to get work out of it that most people don't. Curiosity is what keeps me relevant — I've always got my own projects running, which keeps me sharp on whatever's next and ready to take on more. Custom builds, fractional CTO engagements for founders trying not to burn through their raise, or something I haven't seen before. If it matters to you, it matters to me.
Decades in, still shipping every day. Still curious. Still learning whatever's next. Fortunate to have been part of most of it — and still at the keyboard.
* Bandwidth Available
Still taking on work. If it matters to you, it matters to me. Describe what you're trying to build in plain language — I'll write back within a business day. A real response, not a brochure.