What this looks like in practice
Investment research platform
Screening, reporting, and connected research tools that analysts access through their existing workflow.
Read the case study →Production ML assessment
A review of a production ML workflow that resulted in a sequenced implementation plan.
Read the case study →Document generation system
A document system combining AI-assisted extraction, repeatable assembly, and testing against expert examples.
Read the case study →How it works
Most clients work with me on retainer. Some start with a scoped project.
Trace One Workflow
Start with the people doing the work. Follow one real process end to end. Decide what is worth building, what should change, and what should be left alone.
Put It in Front of Users
Build a working version early. Use feedback from the people doing the job to find what works and what does not.
Deploy and Hand Off
Put the useful version into production. Document it and leave your team able to maintain and extend it.
Services
My main work is embedded AI engineering. Assessments, coaching, and workshops are smaller ways to start.
Embedded AI Engineering
First I help determine what to build. Then I build it. If you have engineers, I work alongside them and leave them able to carry it forward. If you don't, I handle the technical side: architecture, code, deployment, and documentation.
- AI systems that carry out multi-step work
- Data pipelines and putting ML models into live products
- Testing systems that measure whether AI output is correct and useful
- Dashboards and internal tools
- Connecting AI to your existing tools and workflows
- Coaching engineering teams that use AI coding tools
Discovery & AI Assessment
A focused review of a workflow, an AI idea, or an existing ML system before you commit to a larger build. The result is a written recommendation: what to change, what to build, and what to leave alone. It can stand on its own or lead into implementation.
- Interviews with the people doing the work
- What to build, what processes to change, what to leave alone
- Specific next steps, in writing
AI Development Coaching
I work with your engineering team over weeks to set up the docs, workflows, and conventions that make AI coding tools useful.
- Get your team using AI tools consistently
- Project docs and conventions that give AI real context
- Ongoing coaching until the team doesn't need me
Workshops & Talks
Working sessions for teams figuring out where AI fits. In person or remote.
- Curriculum built from your team's real work, not a stock deck
- Hands-on exercises, follow-up materials
- For technical and non-technical rooms
Writing and conversations from the work
Five Lessons from Putting AI Into Research Teams
Which tasks to automate, where analysts push back, and how to tell if a workflow is quietly getting worse. From a talk I gave to a private group of finance professionals.
Read →Four Building Blocks for Document Generation Agents
The patterns I use when building agents that generate documents from messy source material.
Read →Invest with AI: Deploying AI on the Buyside
I joined Brett Caughran and Khe Hy for a practical conversation about choosing one workflow, organizing firm knowledge, evaluating AI output, and getting from prototype to production.
Listen →
Matt Stockton
I've built software for more than 20 years and worked in ML and AI for roughly a decade. I started inside hedge funds building trading systems, then built ML platforms and high-volume data systems at startups.
Now I run PragmaNexus. Recent client work spans investment research, document systems, production ML, logistics, and AI-assisted development. I work inside client teams, sometimes alongside their engineers and sometimes as the technical operation.
I write about the work at mattstockton.com.
Common Questions
Do we need an internal engineering team?
No. I can work alongside your engineers or handle the technical work myself. If your team will own the system, I document it and involve them early enough that the handoff is real.
What if AI isn't the right solution?
I'll tell you. Sometimes the answer is changing how the work gets done — a spreadsheet and a process change can beat any AI you'd build for it. I've talked people out of building when that was the case.
What if I just need help thinking through AI?
Some engagements are assessment only. I review the workflow or system and deliver a written plan your team can execute. A larger build is not assumed.
What does the engagement look like?
Most build engagements are monthly retainers. Assessments are usually scoped projects lasting a few weeks. Workshops are single sessions. We choose the smallest structure that fits the problem.
Tell me what you're working on
Tell me what is stuck, what you have tried, and what a useful outcome would be.
Prefer email? Reach out directly:
[email protected]A few sentences about what you're dealing with is plenty.