Guacamole

Product Logo
Architecture Diagram
Products Used
Analytics
Analytics 2
Checkpoints

Inspiration

Food pantries run on thin margins, volunteer time, and paper or ad hoc spreadsheets. Staff need to know what’s on the shelf now, move stock between programs quickly, and understand trends without a data team. We built Guacamole so volunteers can log and look up inventory with minimal friction, and so managers get trustworthy analytics without standing up a separate BI stack. The goal is simple: less admin overhead, fewer stockouts, faster decisions.

What it does

Guacamole is an AI-powered food pantry app for volunteers to track inventory, run fast lookups, and see real-time analytics.

Computer vision product logging streamlines how items get into the system. Real-time stock keeping keeps counts aligned with what’s actually on the floor. Inventory analysis turns movement and history into something managers can act on. A checkpoint baseline system helps establish and compare expected vs actual stock over time. Volunteer management stays lightweight: invite codes and broadcasting so onboarding and coordination don’t become a second job.

Together, that’s a single flow from capture → stock → insight → coordination for pantry operations.

How we built it

Frontend

React Native with Expo — one codebase for mobile and web-style deployment. Expo Router — file-based routing for clear navigation between volunteer and manager flows. TypeScript — safer refactors as features grew. OAuth2 SSO — familiar sign-in for organizations already on Google / identity providers.

Backend & data

Firebase Authentication — email/password plus Google and other SSO paths. Cloud Functions — heavier work off the device: BigQuery jobs, Gemini / AI calls, and other server-side orchestration. Firestore — live inventory and app state. Firebase Storage — photos and file-backed logging for products and audits. Analytics & AI

Google BigQuery — warehouse for historical inventory and analytics-friendly OLAP-style queries on top of operational data. Vertex AI — hosting and routing for Gemini (and related) model workloads.

Challenges we ran into

Split roles — Volunteers need a dead-simple path; managers need editing, analytics, and guardrails. Designing one app with two mental models took iteration. Real-time + history — Firestore excels at live updates; long-running trends belong in BigQuery. Wiring both without duplicate sources of truth was a design puzzle. AI and CV in the wild — Lighting, packaging, and noisy backgrounds make product recognition uneven; we had to balance automation with manual correction. Mobile UX at scale — Charts, modals, and offline-ish behavior on variable networks forced us to tighten loading states and error handling. Security and invites — Invite codes and role separation had to stay understandable for non-technical pantry staff.

Accomplishments that we're proud of

A unified volunteer + manager experience on React Native / Expo with clear navigation via Expo Router. An end-to-end pipeline: capture and log items, sync inventory in Firestore, and push historical analytics into BigQuery. Server-side intelligence in Cloud Functions so phones stay fast while Gemini and heavy queries run in the cloud. Volunteer management that fits real pantries: invites and broadcasting without enterprise IT overhead. Shipping something that could plausibly replace spreadsheets for day-to-day pantry ops—not just a demo.

What we learned

Firestore and BigQuery complement each other — operational truth vs analytical history; the trick is clear ownership of each dataset. File-based routing (Expo Router) pays off when the app splits into volunteer vs manager surfaces. AI features are only as good as the fallback UX when the model is wrong—always leave a fast manual path. Pantry users are time-poor; every extra tap on the volunteer path is a real cost during distribution hours.

What's next for Guacamole

Deeper forecasting (seasonality, donation patterns) on top of BigQuery. Stronger offline-first flows for warehouses with spotty connectivity. Multi-pantry / regional dashboards for food banks that support many sites. Tighter audit trails and exports for funders and compliance. More CV and NLP assist (e.g., bulk receipt or label capture) with human-in-the-loop review.