I combine hands-on AI development, HCI research, and design strategy to turn undefined problems into things teams can build.
A LangGraph-based multi-agent system for consulting sales experts, built solo in 5 weeks. I owned it end to end, from product definition through evaluation and deployment, and drew the line between what the agents automate and what stays personal.
An AI diagnostic tool for thyroid pathology that had never met a pathologist. I reframed the work from feature validation to trust validation: does this earn a place in the workflow?
A 300-person consultancy whose expertise lived in people, not systems. I designed and ran the workshop that decided what got built, and surfaced the dependency everything else hinged on.
Strategic consulting for the U.S. Department of Education: researching and building first-of-their-kind data visualization prototypes for government education testing, and using AI to surface prediction uncertainty over longitudinal data.
JavaScript
Python
React
D3.js
SQL
LangGraph
Gemini API
Claude Code
Cursor
GitHub
Vercel
ngrok
Supabase
Hugging Face
Ollama
Next.js
The fastest way to understand a new technology is to make it do something useful. Independent tools and experiments I shipped on my own, usually to answer a question no off-the-shelf product would.
A live experiment testing whether different interfaces over identical data actually improve decision quality.
A 30+ column dataset (telemetry, surveys, coded responses) consolidated into one view: distributions, correlations, t-tests, and more.
| Comparison | Δ accuracy | t | p |
|---|---|---|---|
| Sandbox vs table | +0.42 | 2.31 | .025* |
| AI chat vs table | +0.31 | 1.94 | .058 |
| Scrollytelling vs table | +0.48 | 2.67 | .010* |
Hear how a stakeholder group would answer, grounded in real interviews, fully local.
A Chrome extension that turns ten minutes of copy-pasting into a one-click fit score.
I grew up in Napoli, Italy, and later moved to the United States. Since then I have earned a Bachelor of Science in Computer Science and Philosophy, and a Master of Science in Human-Computer Interaction. Every problem I care about sits somewhere between people and the technologies that serve them.
Before I was anything else, I spent seventeen years as a classical pianist, which taught me a lesson that transferred to everything else: the unglamorous daily work is what truly matters. When I'm not sitting in front of a piano, I like to read novels (preferably on a beach), lose my voice at R&B concerts, chase a ball around a squash court, and build small utilities nobody asked for. That last instinct turns out to be useful more often than it has any right to be.
I spent the last few years teaching computer science, documenting workflows, running research studies, shipping MVPs, and working across product strategy, design research, and engineering. The common denominator? Figure out the confusing bit, then hand people a solution and see if it works.
Say helloThe philosophy half of my degree never wore off: I still think an argument you can't articulate is just a vibe. These pieces take a position and take their chances.
Intelligence never lived only in our skull. It evolved with our tools, and it has just crossed a threshold no individual could have ever crossed alone. On what machines contribute, what judgment refuses to cede, and why work will soon be measured by it.
By Angela Petrone · 14 min read · Continue reading →Synthetic users promise research insights minus people. But what happens to knowledge when the data we use becomes a copy of a copy of a copy...?
By Angela Petrone · 9 min readWhy surfacing every detail backfires, and how we earned trust with the right information at the right moment in their workflow.
By Angela Petrone · 7 min readWhen one person can ship custom infrastructure in a weekend, the old reasons to buy off-the-shelf stop holding.
By Angela Petrone · 5 min readEngagement and comprehension metrics miss what matters most: whether someone can act on what they saw. Evaluating data products on decision quality.
By Angela Petrone · 6 min readdef translate(team):
# clinicians, engineers,
# executives, agencies
shared = find_common_terms()
aligned = agree(team, shared)
return ship(aligned)Facilitating the conversations that find shared language, so everyone is heard and the team works towards the same thing.
def build(question):
tool = find_tool(question)
if tool is None:
# nothing fits? make it
tool = build_my_own()
return tool.answer(question)When no tool fits the question, I make one, so I can test it and talk to engineers with context.
def scope(problem):
ideas = run_workshop()
vital = prioritize(ideas)
cut(everything_else)
# smaller, but shippable
return tractable(vital)Run the workshop, hear the voices, prioritize what comes up, so time and resources go where they matter.
I’m open to full-time roles starting January 2027,
and I’m always free for a coffee.
I'm Angela Petrone's AI assistant. Ask me anything about her work, her thinking, or how she builds, and I'll answer the way she would.
Angela is a product builder and researcher who takes problems nobody has scoped yet and turns them into things teams can build, working end-to-end across engineering, design, research, and strategy.
Right now she's an AI Product Builder at GlobalLogic, where she shipped a multi-agent sales system solo in 5 weeks. She's validated clinical AI at UChicago Medicine, shipped AI knowledge management at Method, and shaped a federal data product at Education First.
Between projects she builds her own tools: live experiments, local RAG pipelines, Chrome extensions.
Reach her at apetronedesign@gmail.com, LinkedIn ↗.