We went with (Talent Readiness & Skills Intelligence Platform)
Inspiration
Companies lose critical operational knowledge the moment a single expert walks out — and most workforce tools are static dashboards that never actually calculate that risk. We wanted to build something a skeptical operations leader would trust: real exposure, in dollars and days, calculated before it becomes a crisis, not discovered after.
What it does
Foresight calculates live workforce readiness, critical skill gaps, single-point-of-failure risk, and financial exposure from evidence-weighted employee skill data — nothing is hardcoded. It lets you simulate "what if this person left today" and see readiness and cost-of-disruption recalculate in real time. It ranks internal backup candidates for any skill by fit, growth trajectory, and career interest, then auto-generates a specific 30/60/90-day development plan per employee. Two Google Gemini-powered features follow a human-in-the-loop model — AI proposes a skill-score update from new evidence, or translates a written business strategy into future skill targets — but a manager has to approve before anything actually changes. It also accepts real CSV/Excel uploads of a company's own workforce data.
How we built it
FastAPI + Python backend built around a deterministic, unit-tested calculation engine — every headline number traces back to a documented formula and named thresholds. Frontend is a single-file vanilla JavaScript SPA, no framework or build step. Google Gemini powers the two live AI features, chosen specifically for its genuine no-credit-card free tier. Continuous deployment to Vercel straight from GitHub.
Challenges we ran into
Keeping every number mathematically consistent as we expanded from an initial small roster to 40 realistic employees — a single formula change anywhere had to be checked against every page that displayed it. Building a real empty state so a brand-new account doesn't look suspiciously pre-loaded with fake data, which we knew a judge would ask about directly. And working around Gemini's tight free-tier rate limits: we hit quota exhaustion ourselves during testing, so we built retry logic and a second-key failover so an AI feature fails gracefully with an honest message instead of breaking the demo outright.
Accomplishments that we're proud of
A calculation engine covered by automated tests, not just a demo that happens to work once. Every number on every page — readiness, gaps, risk, dollar exposure — is provably derived from the same underlying employee dataset, so changing one employee's evidence moves the whole picture consistently instead of just one screen. And a real file-upload pipeline that actually parses and imports CSV/Excel data, not just a UI mockup of one.
What we learned
(worth writing in your own words — but one real thing that happened: we ran straight into the same "single point of failure" problem our own product is designed to detect, when our AI features all shared one Gemini API key. Building the fix — a second key, graceful fallback — was a pretty direct, lived example of exactly the kind of risk Foresight is built to surface for a company's workforce.)
What's next for Foresight
Real persistent storage — Postgres via Supabase — so a company's uploaded workforce data is actually saved and shared across sessions, instead of living only in one browser tab until it's refreshed.
Log in or sign up for Devpost to join the conversation.