OptiGraph

OptiGraph – Project Story

🌟 Inspiration

I’m obsessed with Moneyball: if sabermetrics can revolutionize baseball, why not business operations?
OptiGraph applies that same lens and process using statistics and mathematical analysis to KPIs—revealing hidden levers in sales, supply-chain, or finance data.

📚 What I Learned

Even simple ratios (revenue per employee) still make executives say “wow.”
4-sentence vector chunks hit the sweet spot for RAG latency and relevance.
Game-theory payoff matrices are surprisingly intuitive for pricing strategy.

🛠️ How I Built It

Layer	Tech & Flow
Data Intake	One-click CSV upload ➜ stored in Supabase with metadata.
KPI Engine	Python edge worker computes sabermetric ratios & forecasts.
Narrative Insights	OpenAI o4-mini turns numbers into plain-English recommendations.
RAG Chat	Embeddings indexed; users ask follow-ups via a private chat assistant.
UI	Ant Design dashboard: file stats, ROI cards, and risk flags.
Hosting	Netlify CI/CD; Supabase RLS for data isolation.

⚠️ Challenges

Issue	Fix
Auto-detecting numeric vs categorical columns	Schema sniff + sample-row heuristics.
Netlify 25-second limit on 50 MB uploads	Streamed chunks to Supabase Storage, then processed async.
Explaining scary outliers diplomatically	Added “Insight Tone” layer that softens language when variance >2σ.