I’m not the person who hand-writes 50,000 lines of C++ from memory — I’m the person who knows how to direct AI to do it, with the judgment to steer, verify, and ship the result. Give me a real problem and I’ll turn it into a working tool. Every project below — this site included — was built by orchestrating AI agents.
Each one I built by directing AI agents end to end — framing the problem, steering the build, verifying the result. The ambition of what’s here is the resume: it’s what becomes possible when you know how to wield the tool. Click in for a live, interactive walkthrough.
PDFs → queryable knowledge graph → autonomous, grounded research agents
A document-intelligence engine that turns PDFs into a queryable knowledge graph, then runs autonomous deep-research agents with epistemic grounding and an LLM pool across a GPU fleet.
Open interactive demo → 🗄️ 02 / personal project · systems engineeringBuilt from scratch in C++17 · 110/110 tests green
A production-grade relational database built from scratch — ACID transactions, a cost-based optimizer, vectorized execution, ARIES recovery, and a TLS SQL wire protocol. 110/110 tests green.
Open interactive demo → 📸 03 / the master group · applied computer visionOne photo → on-the-spot damage detection · active-learning label loop
Snap a photo of a rooftop HVAC unit; a detection model flags damage on the spot. A human-in-the-loop labeling loop ranks corrections by model disagreement to squeeze signal from a hard, sparse dataset.
Open interactive demo → ⚡ 04 / the master group · low-code data platformDeployed company-wide · the pipeline that feeds the RTU model
A company-wide Power Platform app: PowerApps collects 11-point & outbound QC inspections, Power Automate stores, categorizes and cleans the data in SharePoint — the automated pipeline that feeds the RTU QA model.
Open interactive demo →No one assigned me these projects, trained me in these tools, or checked my work. That’s the thread that ties them together: I get curious about a problem, research the domain until I actually understand it, and adapt what I already know to whatever’s new. AI is the ultimate lever for someone who works this way — it turns “I don’t know this yet” into “let’s build it.”
Get curious about how work actually happens, and find the real friction behind the stated one.
Teach myself the concepts and vocabulary until I can reason about it — not just search for it.
Map it onto tools and patterns I already know, then drive AI through the build.
Read the output, test it, catch what’s wrong, and push back until it’s right.
Get it into real hands and write down how it works, so someone else can pick it up.
These projects span databases, computer vision, knowledge graphs, and low-code business apps. I had no formal training in any of them — I picked each one up by researching the domain and adapting what I already knew. When the tools keep changing, that’s really the only skill that lasts.
My fascination with computers started with a game — modding Garry’s Mod as a kid, where I discovered software’s rules were written, and could be rewritten by anyone willing to tinker. That instinct — opening the machine up and asking what it could become — has carried me from game mods to quantitative models to training my own detection transformers, all self-taught, with no traditional CS background. Today I’m a Quality Control Inspector at The Master Group, where I’ve treated the role as a mandate to improve the process itself, not just execute it.
My edge isn’t deep expertise in any one language — it’s knowing how to point AI at a problem and drive it to a real, working result. I’m comfortable with both technical and non-technical teammates, I document as I go, and I care most about tools people actually use. Understanding of code lets me steer and verify what the AI produces, rather than take it on faith.
That instinct comes from a lifetime around computers. I’ve been tinkering with them since I was young — comfortable across Windows, Linux, and macOS, at home on the command line, and hands-on with web development, Python, data analysis and SQL. Add the full Microsoft 365 Suite and Power Platform, and I can meet a problem wherever it lives — from a shell script to a SharePoint list.
Treated a QC role as a process-improvement mandate — designed a Power Apps inspection platform now running in three distribution centres, and now the detection-transformer pipeline it feeds.
Built a from-scratch C++ database engine and an agentic research assistant over my personal library — on my own time, each a test of how far building with AI can go.
From modding Garry’s Mod as a kid to building quantitative financial models — everything learned by staying curious about how systems actually work.
The work I’m proudest of at The Master Group started as conversations on the floor — a workaround someone invented, a frustration voiced three times a week, a “why do we even do it this way?” Those aren’t complaints; they’re project specs waiting for someone who can hear them, define them with data, and orchestrate the tech to solve them. The personal projects here come from the same instinct pointed inward — a “how does this actually work?” I couldn’t leave alone. I’ve been doing this at the edges of my role — I want to do it deliberately, at scale, with a team built for it.