A career crossing hardware, software, and the systems in between.

Role

Vibe Coder

Location

New York, NY

An independent chapter: small products, shipped fast, no roadmap.

• Building StreetEasier and StayCompare, shared workspaces that turn apartment hunting and group-trip planning into structured comparisons instead of scattered tabs and chat threads.
• Shipping browser-first tools on Cloudflare Workers, D1, and R2, and seeing how far that stack can go before a traditional backend earns its keep.
• Building Codexscope, a local app and CLI for searching AI coding history, recovering past decisions, and tracing delegated work across sessions.
• Turned a vintage payphone at a Chinatown pizza parlor into a working, interactive prop. Keypad combos trigger songs and sounds, and a web dashboard shows what people are punching in so the shop can make a familiar combo do something new. Mostly, I wanted an excuse to solder and hack on a Raspberry Pi again.
• Helping re-found Homebrew, a three-floor Manhattan coliving community for founders and builders.

Role

Co-Founder & CTO

Location

New York, NY

Co-founder and CTO of a venture-backed health tech startup that raised a $3M seed. Led product and engineering across telemedicine, patient billing, and senior care through to an acquisition by Commure/Athelas.

• Built the telemedicine platform's FastAPI backend end-to-end: accounts and nurse onboarding, scheduling with tag-driven patient–provider matching, embedded video visits (Zoom, JaaS), and a templated Twilio SMS engine. Kept messaging nurse-driven to build a training dataset before automating.
• Shipped Contour Pay from direct practice feedback: DOB-gated payment links, card and wallet payments, clinic reconciliation, and an AI collections agent (LangChain, then LangGraph, with Gemini) that drafted patient SMS from invoice, balance, and conversation context; every draft passed an operator's review before sending. Piloted with a teledermatology practice.
• Built Contour Care, an AI senior-care search product ("ChatGPT meets Zillow" for nursing homes): streaming agentic chat (Vercel AI SDK, Gemini) driving a tool-calling backend over a 53,000-facility PostGIS database, generative UI, and Trigger.dev deep-research reports, drawing thousands of visitors and 2,700+ placement inquiries.
• Built the data engine behind Contour Care: staged ETL with geocoding failover and Gemini pipelines that generated structured facility profiles across all 50 states.
• Built an outbound voice-agent platform to fill facility data gaps: a timezone-aware dispatcher sent dual-agent squads (Vapi, Deepgram, ElevenLabs) through phone menus to interview staff; post-call extraction wrote structured updates, attempt history, and lineage into our database; live transcripts streamed to ops consoles over Supabase Realtime.

Role

Founding Engineer, Briefly

Location

San Francisco, CA (remote)

Helped friends lay the backend groundwork for Assort's healthcare voice agents.

• Built the initial Python/FastAPI backend, task infrastructure, and containerized GCP/GKE deployment.
• Connected the agents to Epic through Redox and FHIR, letting them check provider schedules, book appointments, and update patient records during live calls.

Role

Senior Software Engineer

Location

Mountain View, CA

Athelas was where roles became suggestions. I worked on the first FDA-cleared at-home CBC device, built the portal nurses used to monitor more than 100,000 patients and escalate to physicians, then embedded with an acquired medical-billing team in India to turn its manual operation into software.

• Built remittance ingestion across APIs, web portals, faxed PDFs, and EDI 835 files. Automatically processed volume grew from $200K in November 2022 to over $2M per month across 70+ practices, while payment-processing errors fell by over 90%.
• Embedded with the billing team in India, mapped how denied claims moved through the operation, and turned manual handoffs into a production line where we could inspect the work, batch it, and gradually automate it. Reverse-engineered a payer clearinghouse backend and built async collectors that checked thousands of claims in bulk, surfaced status and payment discrepancies, and fed the results into the same workflow.
• Led software and firmware for Athelas Home through its FDA clearance. Optical calibration raised manufacturing yield from 60% to over 80% and cut device testing from 12 minutes to 4.
• Co-designed a new self-test strip with the mechanical team and built its visual validation pipeline: deterministic OpenCV checks by default, with an EfficientDet model I trained and deployed through TensorFlow Serving as the ML fallback when OpenCV failed or could not decide. Strip faults fell from 20% to 5%.

Role

Software Engineer II, Money Movement Risk

Location

Palo Alto, CA

Backend engineer on Uber's fraud and payment-risk teams, working on systems that handled millions of transactions a day and screened billions of dollars in payment volume.

• Replaced a nightly chargeback job with a real-time service that compiled evidence and submitted disputes to Adyen and Braintree automatically.
• Built Python, Spark, and SQL pipelines that fed real-time fraud models targeting collusion and fake-restaurant schemes on Uber Eats.
• Led the risk integration for peer-to-peer Uber Cash gifting, from launch planning through production ownership.

Role

Founding Member

Founding member during Phonic's 2019 Hult Prize run. After the UN global finals, the remaining team took Phonic through Y Combinator (W20) and a 2023 acquisition by Infillion.

• Pitched the venture at the United Nations Headquarters in New York to judges including President Bill Clinton, as one of 7 global finalist teams selected from thousands in the 2019 Hult Prize.
• Pivoted with the team from a youth-unemployment concept to a smartphone app for guided audio interviews for market research.

May 2018 — Aug 2018 · Palo Alto, CA

On Uber's Maps Hotspots team, we used rider and driver GPS signals at trip start to infer where pickups actually happened when GPS was unreliable in dense cities. I worked on the data pipeline that aggregated those pickups into preferred pickup and drop-off spots, and on the map tools operations teams used to evaluate them.

Jan 2018 — Apr 2018 · Redwood City, CA

I wrote application firmware for Level's smart deadbolt: door open/close detection, shareable keys that expired, and the low-power behavior that let one CR2 battery last more than a year. In electrical design reviews, I made the case for adding a real-time clock, which cut system power use by 15%.

Sep 2016 — Aug 2017 · Palo Alto, CA

Eight months on the Model 3 program, split across two internships. First, with complete prototypes scarce, I built a low-voltage testbench and CAN/LIN tools that let body-electronics firmware development continue without a car, and wrote and tested firmware for Phone Key, which unlocked the car as your phone approached. Back the next summer on body controls, I automated controller testing across −40°C to +80°C and wrote firmware for the sensorless motor driver behind the main HVAC blower.

Jan 2016 — Apr 2016 · Waterloo, ON

I worked on firmware for an OBD-II telemetry logger used by EV fleets and wrote its 4G modem and GPS receiver drivers. The GPS work cut average cold-start fix time from 27 seconds to under four.

May 2015 — Aug 2015 · Kitchener, ON

At Deloitte's innovation lab, our internship project was a helmet-mounted safety system for miners working underground. The modules relayed telemetry helmet to helmet until it reached a central node.