Software Projects
ManuFlow AI / AI Manufacturing Decision Copilot
A manufacturing decision-support demo for exploring production-order delay risk, operational drivers, and corrective actions. The Next.js application uses API routes and typed services, with optional MongoDB/Mongoose, Gemini, and external model-inference integrations. It remains usable with synthetic demo data and deterministic fallbacks when optional services are unavailable.
Tech Stack
Problem / Challenge
Production teams need a clearer way to bring operational signals, delay risk, contributing factors, and possible corrective actions into one decision-support workflow.
Approach
The project combines a Next.js dashboard, API routes, typed services, risk scoring, and an AI copilot. Optional integrations are isolated behind resilient fallbacks so the deployed demo can continue to show data-aware behavior with synthetic demo data when MongoDB, Gemini, or the external Core ML service is unavailable.
Key Features
- Next.js dashboard with API routes and typed application services
- Manufacturing analytics, delay-risk views, and operational recommendations
- Optional MongoDB/Mongoose persistence and Gemini-assisted copilot responses
- Deterministic fallback behavior for unavailable AI, database, or inference integrations
- Apple Create ML classifier workflow with an external Core ML inference integration design
- CI workflow for frontend build checks and Python validation
Outcome
A deployed, demo-oriented manufacturing decision-support application with documented architecture, CI, and clear integration boundaries. It is presented as a synthetic-data demo rather than a production manufacturing deployment.
What I Learned
- How typed services and API routes can separate a decision-support application into clear responsibilities
- How to design deterministic fallback behavior around optional AI, database, and model-inference services
- How a classifier training workflow can connect to an application through an external inference boundary
- How CI and deployment documentation support a repeatable demo workflow