ForgeAI: Smart ERP for Precision Machining

💡 Inspiration

Traditional machining workshops and precision manufacturing plants often struggle with rigid ERP systems. Production scheduling is done manually on spreadsheets, machine downtime is reactively managed, and quoting complex multi-step custom parts takes days.

We built ForgeAI to bridge the gap between traditional metal fabrication and modern cognitive AI. By integrating OpenAI's advanced models, we transform the ERP from a passive "system of record" into an active, intelligent "co-pilot" for workshop owners and floor managers.

⚙️ What it does

ForgeAI is an intelligent ERP designed specifically for precision machining enterprises:

• Instant Smart Quoting: Upload raw technical requirements, customer emails, or text descriptions of custom parts, and ForgeAI automatically parses them to estimate material costs and processing times.
• Dynamic AI Scheduling: Automatically optimizes machining sequences on the shop floor based on real-time machine status, operator availability, and delivery deadlines.
• Predictive Maintenance Alerts: Detects potential equipment anomalies and recommends proactive maintenance schedules using lightweight telemetry logs.
• Conversational Shop Floor Assistant: Machinists can talk directly to the ERP via voice or chat to log completed parts, report machine faults, or request raw materials.

🛠️ How we built it

• Backend Core: Built using Python and FastAPI to handle high-concurrency manufacturing workflows.
• AI Engine: Powered by OpenAI GPT-4o (for parsing complex, unstructured order requirements) and GPT-4o-mini (for low-latency, cost-effective shop floor voice/text interactions).
• AI Agents: Implemented using a multi-agent framework to handle task delegation—one agent manages scheduling, another manages inventory routing, and a third oversees predictive alerts.
• Frontend: A clean, responsive dashboard built with React and Tailwind CSS, optimized for both desktop and workshop tablets.

🚧 Challenges we faced

• Translating Machining Jargon: Aligning natural language queries with highly specific manufacturing/metalworking terminology (e.g., CNC milling speeds, tolerance specs, anodizing post-treatments). We solved this by creating a retrieval-augmented generation (RAG) pipeline packed with machining dictionaries.
• Real-time Agent Synchronization: Ensuring multiple specialized agents could collaborate without creating loops or resource lockups during intense scheduling conflicts.

🎉 Accomplishments that we're proud of

• Successfully developed a prototype that reduces the manual quoting process for custom machined parts from 3 days to under 5 minutes.
• Created an incredibly intuitive voice interface that machinists on the noisy shop floor can easily interact with.

🧠 What we learned

• Large Language Models (LLMs) are exceptionally good at structured data extraction from unstructured industrial RFQs (Requests for Quotes).
• True industrial AI isn't about replacing physical machines, but about maximizing their uptime and orchestrating human expertise.

🚀 What's next for ForgeAI

• CAD/Blueprints Integration: Integrating vision models (like GPT-4o) to directly analyze PDF blueprints and 2D engineering drawings to automate physical dimension calculations.
• On-premise Deployment: Offering hybrid hosting options for enterprises with strict data privacy guidelines.

Built With

• fastapi
• gpt-5.6
• langchain
• openai
• postgresql
• python
• react