Cornell UniversityNew YorkOpen to Product / AI opportunities
Selected work
Ideas turned into systems people can actually use.
One main body of work at Dreame with two branches I built myself, the two products I am building now, and the engineering that came before it. Each one the same way: problem, insight, experiment, result.
Nexus Gen is a privacy first AI NAS: local models, agents, and storage in one box. I joined the 0 to 1 build inside Dreame's incubator, wrote the early PRD and twelve month roadmap, ran the user research, and shipped two of its surfaces with my own hands. The product won the CES Asia Best Innovation Award.
A strong NAS, a strong hardware story, and an open question: what does local AI actually do for the person who owns the box?
02Insight
Nobody wanted AI features. They wanted friction removed from media, files, photos and privacy. The product had to disappear into an existing habit.
03Experiment
Wrote the PRD and twelve month roadmap, ran research with 100 NAS users, moved the agent onto an MCP tool registry, and took a Home Theater beta to 80 users through three iterations.
04Result
78% activation, 81% active daily, 60% D14. Tool selection tokens down 80%, latency down three minutes, timeouts gone. CES Asia Best Innovation Award.
01 · Branch 01
The Home Theater user journey
One evening with the beta, stage by stage. Grey is what the NAS did before. Black is what shipped.
TapQ is an open source macOS app that turns the AirPods people already wear into a private control surface for AI agents. Nod to approve, shake to decline, tilt to choose, speak to follow up. Launched from 50 beta users to 200+ downloads and 500+ GitHub stars with 40% D30 retention, and now expanding from developer agents into Business, Lecture, and Travel modes.
0 to 1Open sourceAirPodsIMUVoiceAgentsRoadmapUser research
500+GitHub stars
200+Downloads, from 50 beta users
40%D30 retention
5Agent adapters: Claude Code, Codex, Cursor, Antigravity, OpenCode
Agents work alone for minutes, then stall on a permission prompt until you walk back to the terminal. In meetings, the moment you need help is the moment you cannot pick up a phone.
02Insight
Use the interface people already wear. In interviews, wake words and voice attribution were the real frictions, and the IMU inside the earbuds can solve both.
03Experiment
Shipped an open source macOS app: nod to approve, shake to decline, tilt to choose, and full duplex voice with IMU voice attribution. Started with 50 beta users.
04Result
200+ downloads, 500+ GitHub stars, 40% D30 retention. Roadmap extended from developer agents into Business, Lecture and Travel modes.
03 · Personal product · Consumer · AI · Live prototype
Choosing a college by its 40 year return, not its ranking.
PathWise turns a ten question intake into a personalized college and career report: net price after aid, entry and year ten wages, a 40 year NPV for every school on the list, and an AI narrative that explains the verdict. Built end to end and running as a live prototype.
0 to 1ConsumerFinancial modelReactAI narrativeData viz
Turning repetitive infrastructure work into software.
Owned the roadmap for an internal GIS/CAD network design tool used by 200+ engineers, and turned repeated operational bottlenecks into automation adopted company wide.
Python tooling, validation logic, one standard workflow
60%faster signal verification
↓
03 · Better operating system
Faster delivery, improved accuracy, less engineering time lost.
01Problem
200+ engineers designed networks in a GIS/CAD tool, but recalculation, BOM pulls and error checks were all done by hand around it.
02Insight
Interviews with 30+ designers kept naming the same two time sinks: recalculating signal levels after a change, and assembling bills of materials.
03Experiment
A three click Python automation for tracing, recalculation and verification, and an Auto BOM that records equipment changes as they happen. One team first, then three regions.
04Result
Recalculation 90% faster, manual errors down 25%, design turnaround 40% faster, Auto BOM used company wide, about $700K saved per regional operation.
The industries changed: plasma physics research, electrical equipment engineering, telecommunications infrastructure, software automation, AI products. The underlying habit did not. Understand the system, find the bottleneck, build a better solution.
I like complicated systems. I like making them feel simple.
My work has moved between physical infrastructure, software, AI, engineering, and product. What interests me most is usually the same problem: understanding how a complex system works, finding where users experience friction, and building something better.
Experience
Engineering depth. Product direction.
I did not start by trying to become a product person. I started by finding problems I wanted to fix.
2026 to Present
Cornell University
M.Eng, Systems Engineering
Graduate study focused on AI systems and product oriented problem solving, with a research focus on AI systems and open source bug report evaluations.
May 2026 to Present
TapQ (tapq.ai)
Founder and Product Manager
Open source AirPods interface for hands free AI agent control. Head gestures and IMU based voice attribution for full duplex agent interaction; 50 beta users to 200+ downloads and 500+ GitHub stars, 40% D30 retention. Roadmap from developer agents into Business, Lecture, and Travel modes.
Apr 2026 to Jul 2026
Dreame Technology (CES Asia Innovation Award)
Product Manager Intern, AI NAS Business Unit
Consumer electronics company building privacy first, AI powered home storage devices (seed round, $2M). 0 to 1 development inside a company wide incubator: user research with 100 NAS users, model selection across 7 LLMs, the MCP based tool registry, the dual volume privacy architecture, and the Home Theater MVP that reached 78% beta activation, 81% daily active, and 60% D14 retention.
Apr 2023 to Mar 2026
Spectrum
Designer Engineer III, Product Owner of the internal GIS/CAD design tool
Owned the roadmap for the internal GIS and CAD network design tool used by 200+ engineers, aligning Design, Construction, Operations, and Engineering. Defined the 3 click signal recalculation automation (90% faster, 25% fewer errors, about 2 engineering hours saved per project) and launched Auto BOM (50% faster reports, about $700K in savings per regional operation). Delivered 150+ FTTH and 620+ enterprise fiber solutions across Manhattan.
May 2022 to Mar 2023
Decom Electrical Co., Ltd.
R&D / Project Engineer
Co-engineered a 600mm-wide SF6 gas-insulated switchgear system: 3D models and assembly drawings from concept through prototype, with DFM/DFA review that cut assembly errors 15%.
Jan 2021 to Dec 2022
NYU Tandon, Plasma Physics Lab
Research Assistant
Designed Paschen Curve experiments on gas-discharge breakdown voltage and built a plasma physics lab from an empty facility to operational readiness.
May 2019 to Nov 2019
A.T. Kearney
Consulting Intern
Analyzed customer and store performance data for Nike Factory Stores in China, identifying demand patterns and commercial opportunities; recommendations presented to Nike China executives contributed to a 15% sales increase.
How I work
Three principles I keep coming back to.
01
Start with the user.
Technology matters only when it solves something people actually care about.
02
Understand the system.
Good product decisions come from understanding technical constraints, incentives, workflows, and edge cases.
03
Build, test, iterate.
I prefer prototypes, experiments, and real user feedback over endless speculation.
About
Technical enough to go deep. Product minded enough to zoom out.
Currently studying Systems Engineering at Cornell, I’m interested in product roles at the intersection of AI, infrastructure, developer tools, and intelligent consumer products.
Before Cornell, I worked on telecommunications infrastructure in New York, where I started building automation tools for problems I kept running into in my own workflows. That pushed me deeper into product development: talking to users, defining problems, prototyping solutions, evaluating AI systems, and working across engineering and business.
When I’m not building things, you’ll probably find me playing golf, exploring new technology, or spending too much time analyzing markets.
Product
Product Strategy · Roadmaps · PRDs · User Research · Prioritization · Experimentation · Metrics · Cross-functional Leadership
AI
LLMs · AI Agents · MCP · Tool Calling · Model Evaluation · Prompt Engineering · Local AI
Technical
Python · SQL · APIs · Automation · Networking · Data Analysis · Systems Engineering · R
Tools
Figma · Canva · JIRA · Git · SolidWorks · AutoCAD
Education
Cornell University
M.Eng, Systems Engineering
Expected Dec 2027 · Research focus: AI systems and open source bug report evaluations
New York University, Tandon School of Engineering
B.S., Applied Physics
Sept 2018 to Dec 2022
Contact me
Let’s build something interesting.
I’m always interested in conversations about AI products, product management, infrastructure, and ambitious technical ideas.
