Jonathan Gu

Jonathan Gu

I build software systems that learn from use.

Right now my main focus is Bountiful Garden, a neighborhood harvest-sharing network that turns backyard abundance into real community. I'm also building OpenClawBrain, a native OpenClaw context-engine plugin that combines lossless transcript memory, a learned routing graph, replay-gated packs, and decision traces. Pelican is a private research system for autonomous options trading.

OpenClawBrain front doors: site · GitHub repo

ML Engineer at Instacart. PhD Economics, UCLA. BA Statistics & Economics, UC Berkeley.

See What I'm Building Get in Touch
What I'm Building
Bountiful Garden neighborhood produce sharing
Live

Bountiful Garden

Your neighbor's Meyer lemon tree dropped 200 lemons last week. You want six. Bountiful is the bridge — post your harvest, browse what's ripe nearby, and grab it off the porch. A farmers' market that never closes, where every stall is someone on your block.

Local-first Full-Stack React
OpenClawBrain memory graph visualization
V2 — live

OpenClawBrain

Standard retrieval starts with amnesia every session. Brain keeps a scorecard, re-shelves the library after every run, and compounds what it learns. V2 ships a native policy-gradient learning path: 344 trajectories traced end-to-end, every route scored, decisions reconstructed from real supervision. The trained pack is published and verifiable.

npx @openclawbrain/cli@0.4.10 learn materialized/promoted pack-d099f991
Product Site 2016 Paper
Project Pelican autonomous options trading
Private R&D

Project Pelican

A research system for autonomous options trading. Pelican watches the market, builds its own features, trains fresh models, and sizes positions — all without a human in the loop. Private infrastructure; notes available on request.

Quantitative Finance ML Pipeline Autonomous Systems
About

Instacart

San Francisco

Machine Learning Engineer

At Instacart I build production ML systems that make real-time decisions at scale across inventory, promotions, and ads. Before that I did a PhD in economics at UCLA, where I spent years thinking about incentives, learning, and dynamic systems.

  • Built and maintain the batch and real-time inventory-availability platform — production modeling, threshold calibration, and predictions across millions of SKUs.
  • Built automated bidding and auctioneer systems for the Ads platform, allocating spend in real time across a live auction.
  • PhD Economics — UCLA, 2014–2020
  • BA Statistics & Economics — UC Berkeley, 2007–2011
Research
Contact

I'm especially interested in conversations about agent infrastructure, applied ML systems, and local-first product design.

Email Twitter / X LinkedIn GitHub Resume (PDF)