I figure out where AI fits your work. Then I build it.

I've spent 20 years building software, much of it at investment firms and fintech companies. Now I build AI tools, mostly for investment research teams. I work inside your team on the strategy and the code.

See the work

At an investment firm? Start here

What I build

All six case studies →

Coding agents for engineering teams

Shared instructions, skills, and architecture docs that give every engineer's agents the same context, plus coaching in your own codebase.

Example: coding agents across a team

How I work

  1. 1

    Learn the work and write it down

    I sit with the people doing the work and write down how it gets done. Most of the job is getting that right.

  2. 2

    Build one thing with the people who use it

    We pick one workflow and build the smallest version that helps. The people who do the work try each version, and we adjust as we go.

  3. 3

    Ship it and keep checking it

    I put it into production with tests built from work your team has already done, so you'll know if it gets worse.

Ways to work together

Monthly retainer · most of my work

Building with your team

First we figure out what to build. Then I build it: architecture, code, deployment, and documentation.

A few weeks, fixed scope

Assessment

A review of one workflow, an AI idea, or an existing ML system, before you commit to a larger build. You get a written plan.

Example: production ML assessment

Weeks to months

Coaching for engineers

I work with your engineers in their own codebases on how they use coding agents, and build the shared context those agents read.

Single session

Workshops and talks

Working sessions built from your team's work, not a stock deck. In person or remote.

Example: applied AI workshop

Writing

All posts at mattstockton.com →

About

Before PragmaNexus I built trading systems, investment platforms, and the data infrastructure underneath them. About a decade of that was ML and AI.

Now I run PragmaNexus. Even the advisory work is hands-on, in the codebase.

I write about the work at mattstockton.com. Full background

  1. 2021–24
    Teamshares · Director of Engineering

    First data leader. Built the data infrastructure that ingested financial data from nearly 100 businesses, and hired the first data team.

  2. 2020–21
    Racon Capital Partners · Director of Data Science and Engineering

    Sole engineer for a quantitative macro strategy: terabytes of financial data, model orchestration, and portfolio risk tools.

  3. 2014–20
    CircleUp · Director of Engineering

    Built the engineering team and the platform that sourced and scored consumer companies for investment. It grew to analyze more than 100,000 businesses.

  4. 2010–14
    Harqen · VP of Engineering

    Four of us ran the whole stack for a video and voice interviewing platform. Millions of applicants used it.

  5. 2007–10
    Stark Investments · Senior Software Engineer

    Derivatives and credit default swap trading software for a $14 billion hedge fund, through the 2008 financial crisis.

  6. Earlier

    Security software for network-connected medical devices at GE.

Questions

Do we need an internal engineering team?

No. If you have engineers, I work alongside them. If you don't, I cover the technical side and write down how it works.

What if AI isn't the right solution?

Sometimes the answer is an AI system. Sometimes it's ordinary software, a process change, or a decision not to build anything. I'll say so when that's the case.

Contact[email protected]

Tell me what you're working on

What's stuck, what you've tried, and what would help. A few sentences is plenty.

Or email