A career inside one industry.
I've spent my career inside investment management and watched the technology underneath it transform four times. Mainframes to client-server to web to cloud to AI. Each transition came with new vocabulary; the underlying business has stayed remarkably consistent.
The starting point is 1983, when I was thirteen, selling and servicing computers at a small shop in San Diego. I worked through high school doing it — Gompers Magnet for computers, science, and math — taking formal coursework in BASIC, FORTRAN, Pascal, and assembly with compiler design. The senior project, which I still remember more vividly than most things from my twenties, was a team build of an original programming language we called SIMPLE, with a Pascal-based compiler the students wrote ourselves to translate it into assembly. Saturdays of those years went to regional programming competitions. I never became a professional developer, but those years are how I learned to think about systems at the level of language and instruction rather than only frameworks.
By twenty I co-owned a computer consulting and retail firm that grew to a million-and-a-half in annual revenue, working on the sales and systems-engineering side. We sold systems to Hughes Aircraft, CSC, General Dynamics, the U.S. Navy. Two thousand computers to the Russian government, around the time the Berlin Wall came down. From there, two years at ComputerLand of San Diego as a Senior Systems Engineer — voted Systems Engineer of the Year both years, posting record billings.
In 1996 I joined Duncan-Hurst Capital Management as their first IT hire. Eighteen months later I was IT Director, reporting to the CEO; by the end I was VP of Information Technology and a Limited Partner. The firm grew from $1B to $8B in assets under management while I was there. Over thirteen years I co-built a real-time portfolio performance and attribution system that financial-systems vendors tried to buy from us, ran the firm's first disaster-recovery program, and managed Y2K.
As the firm wound down and headcount shrank, Beau Duncan — the majority owner and CEO — began giving me progressively more of the firm to run. I took on compliance administration. I became the backup trader when our main trader was on PTO. I worked on the financials and served as the day-to-day interface with outside legal counsel. By the end I had sat in nearly every chair on the buy-side: operations, trading, compliance, finance, legal, technology. The CEO's phrase that stuck was: “Give it to Jarl, he gets things done.”
After Duncan-Hurst closed I spent four years as Senior Business Systems Analyst at Relational Investors, an $8 billion activist equity firm. I implemented the firm's first automated trade order management system, built out our first disaster-recovery site — and was called on to invoke a real recovery when a regional power outage took the office down. The DR test wasn't a test that day.
A few years in, the firm also started having me serve as backup to the head trader when he was away. The trading itself was mechanical — orders routed through VWAP algorithms — but the responsibility was real: an eight-billion-dollar activist fund's trading desk, on my watch.
In 2014 I joined INDATA, the software firm whose OMS I had selected for Duncan-Hurst years earlier. Twelve years in, I'm still here as Director of Product Management, reporting directly to the active owner and cofounder. I lead a team of ten developers — backlog, product roadmap, stakeholder interface — and I'm the sole merge authority on the Nexus repositories. Inside the firm I'm the AI-tooling champion: I rolled out GitHub Copilot to the engineering team, brought in Claude Team accounts for Claude Code, and run regular internal sessions on shipping with AI assistance.
The work I'm proudest of is the most recent. From February 2025 through May 2026 I designed and led the development of INDATA Nexus™, our AI Platform for buy-side firms, working alone with Claude Code as my development partner. It launched publicly in May 2026 and is in production at one client today, with more rollouts in progress. A new “AI-forward” developer is about to join the team to help extend it.
Between 2024 and 2026 I worked through three of Edward Donner's curricula on modern AI engineering — LLM application development end-to-end, building real agent systems with LangGraph and the major agent SDKs, and working with agentic IDEs including Claude Code, Antigravity, and MCP tooling. The reading and exercises ran alongside the building, not before it. Continuous learning at fifty-five isn't a virtue at this point; it's the table stakes of working in this field.
None of my forty-year-old programming coursework is current in the way a 2026 software engineer would describe theirs, and I was never a professional developer in that sense. But the foundations haven't faded. I still write complex SQL with multi-table joins, CTEs, and stored procedures without assistance when the work calls for it. I'm comfortable across the full stack — the OSI layers, the web-server tier, Linux hosts, Docker, IIS URL rewrites, SQL Server administration, the difference between application logic and infrastructure plumbing. I built most of Nexus's production infrastructure myself.
It's fair to point out that AI tools have made a lot of this work more accessible to non-specialists, and they have — that's a good thing. An amateur with a good AI assistant can install Ubuntu, get a container running, and feel like they've shipped something. What AI has not yet democratized is the judgment about which of fifteen possible approaches is the right one, what's about to break under production load, and what to do when the suggested fix is subtly wrong. That judgment is what compresses what would otherwise be six months of stumbling into six weeks of shipping. An agentic IDE is a force multiplier when you can evaluate what it just wrote against principles you understood long before AI got involved — and a liability when you can't.
Outside the building I run marathons — eight finishes so far, four of them Abbott World Marathon Majors. Boston this April (qualified by time), Chicago last fall (also qualified). Berlin in September will be number nine. London in 2027 will be number ten — and will complete the 6-Star.
That's the shape of it.