BoxSight builds applied-AI mobile products on a single shared backend — news intelligence, expense tracking, sustainability research, and time-based discovery.
One backend. One auth layer. One deployment pipeline. Four products sharing everything that matters.
Every app runs on the same production stack — PostgreSQL + pgvector, FastAPI, and Redis — with shared authentication and a common deployment pipeline.
A cross-review engine queries five LLMs in parallel to validate AI output and surface disagreement, reducing single-model failure across the ecosystem.
Each product ships as a real iOS and Android app, not a wrapper — published and downloadable from the App Store and Google Play today.
Four shipping apps, each solving a focused problem with applied AI.
AI news briefings. Ingests 30+ curated sources daily and generates structured, high-signal intelligence through vector search and LLM reasoning.
Smart expense tracking. Snap a receipt and AI OCR categorizes it automatically — then visualizes your spending as a living city that grows or struggles with your habits.
Sustainability research, distilled. Aggregates 20+ university-affiliated feeds into weekly AI video briefings, trend detection, researcher profiles, and threaded discussion.
Nearby places by time, not distance. Two taps — a category plus a time radius — return matching spots with AI-generated labels, ready-made situation bundles, and multi-stop trip chaining.
From ingestion to app, every layer is shared across products — one backend, one auth layer, one deployment pipeline.
One JWT with device binding. SSO across apps, per-app entitlements.
PostgreSQL + pgvector with hybrid search and source attribution.
Five models cross-review in parallel. Disagreements surfaced, not hidden.
Real native binaries on both stores. No wrappers, no PWAs.