ARG Insights is led by a senior AI engineer and researcher with experience on both sides: production systems and advanced ML research.
I teach AI at Duke — courses in applied AI, AI products, and software development. Before consulting, I was a Senior ML Engineer at Apple and at several startups, building systems that had to work in production, not just in demos.
My research focuses on AI systems and agents. I've raised over $1.5 million in research grants and presented at NeurIPS and ICML. That background means I understand both what's actually possible with AI and what it takes to make it useful inside a real business.
The goal is never an impressive demo. It's something that works inside a real team, with real constraints, and delivers clear value.
How we work
We stay intentionally small and senior-led.
I grow the team based on what a project needs, but I don't subcontract loosely or farm work offshore. I either write the code myself or personally review every line that goes into a client system.
When a project needs more depth, I bring in exceptional contractors with strong technical backgrounds — people like:
- a PhD student in AI who has worked at NASA and published at ICML
- a master's-level AI engineer from Northwestern
You get the right people for the project, with tight supervision, technical rigor, and direct accountability throughout.
What clients can expect
Clients work with us because they want practical AI help from people who actually understand the technology and can ship.
That means:
- honest thinking about where AI will and won't help
- systems built for real workflows, not demos
- direct communication with senior technical leadership
- a small, high-quality team matched to the scope of the work
- careful deployment instead of hype-driven experimentation
We're not trying to be the biggest consulting shop. The goal is to do high-quality work, stay close to the details, and build things clients can trust.
What we don't do
We don't bolt on a canned system. There are plenty of off-the-shelf automation platforms that promise to solve everything with a few clicks. We don't use them. Canned systems are built for the average case. Your workflows aren't average. We build custom — designed around your actual inputs, your actual constraints, and the specific business result you need.
We don't disappear and come back with a product. Some shops take a brief, go dark for two months, and show up with something you've never seen. That's not how we work. We live in your Slack. We provide daily updates. You see progress as it happens, you can redirect us at any point, and nothing ships that you haven't already had eyes on.
We don't farm out the work. We write every line of code that goes into a client system ourselves. We don't subcontract — not loosely, not at all.
We don't build demos. The goal is never something impressive in a controlled environment. The goal is something that works inside a real team, with real data, under real conditions — and keeps working after we're done.
We don't pretend AI solves everything. We're honest about where automation will help and where it won't. If a workflow isn't a good fit, we'll tell you. We'd rather lose a project than build something that doesn't create real value.