Handoff

Inspiration

The inspiration of Handoff came from the discovery of inefficiencies in the supply chain. Across 15 factory visits, our team noticed that engineers were struggling on translating CAD and BOM files into RFQs (document needed for supplier quotation to order parts) that specify the hardware manufacturing specifications. Thereby, the procurement buyers also sometimes struggle to understand these documents, causing downstream delays in manufacturing time. Our goal with Handoff is to simplify the handoff of the documentation and to reduce friction in the engineering lifecycle process.

What it does

Handoff is an AI powered platform that takes in a single part STEP file of a 3D model and parses its dimensions. It then allows for additional user input before using AI to draft a 4 page, ISO compliant pdf file. Handoff then finds potential, reputable buyers from a specified location, allowing manufacturers and buyers to easily communicate and hand off RFQs. Manufacturers can input a location, and receive several potential buyers in that area that are searching for that

How we built it

We built Handoff as a web application using Python, HTML, and CSS. Python reads the CAD files with CadQuery, extracts part information, organizes engineering requirements into JSON, communicates with the Claude API, and generates the final RFQ document. Claude receives structured CAD data and engineer-provided requirements, translates that technical information into procurement-friendly language, identifies missing or ambiguous details, and drafts fields such as the scope of work, manufacturing notes, inspection requirements, packaging instructions, and questions for suppliers. Claude is instructed to distinguish confirmed engineering data from AI-generated suggestions and to avoid inventing critical specifications. We used HTML to build the application’s interface and organize the engineer’s workflow, including file uploads, requirement forms, generated RFQ fields, warnings, and document previews. CSS allowed us to create a clear, consistent design that makes complex technical information easier to review. OpenAI Codex as a development assistant to help us write, debug, organize, and integrate the application’s code.

Challenges we ran into

One of our largest challenges was translating complex CAD geometry into information that an AI system could interpret safely. We needed to distinguish between facts extracted directly from the CAD model and suggestions generated by AI so that the system would not invent dimensions, materials, tolerances, or manufacturing requirements. We also had to coordinate several data sources and processing stages. The CAD reader, engineering form, AI drafting system, and PDF renderer all needed consistent JSON schemas to communicate correctly. Another challenge was determining what information could be confidently automated and what should be flagged as missing or uncertain. Manufacturing documents can involve strict tolerances, certifications, finishes, and regional compliance requirements, so even a small ambiguity can have expensive consequences.

Accomplishments that we're proud of

We are proud of being able to build a fully functional prototype that meets our expectations and the ISO guidelines in a short time frame. The prototype was able to successfully read and extract information, turn it into a more readable format, help identify missing information, separate the facts with AI-generated suggestions, and produce and professionally formatted documents even with more complex CAD models. We were able to solve a problem we noticed across many factories, and reduced the highly intensive process down to a few clicks.

What we learned

We learned that the main challenge is communicating the project accurately across different disciplines and technical knowledge. We also learned that while AI is incredibly powerful, it should never interfere with the data and facts. CAD measurements and engineer-entered requirements should remain the authoritative sources, while AI should organize the information, improve clarity, identify omissions, and generate questions. Our factory research showed us that RFQ generation is only one part of a larger communication problem spanning engineering, procurement, suppliers, compliance, and manufacturing. Reducing friction at this Handoff could save engineering time while also preventing expensive downstream mistakes.

What's next for Handoff

We plan on launching design partnerships with our industry connections, beginning with small to mid sized manufacturing companies’ engineering departments in the USA and India. Through these partnerships, we hope to integrate their feedback to improve our product and analyze what features within the product are most valuable to reducing friction in the RFQ generation and engineering lifecycle process. To expand Handoff, we plan to offer free trial and pilot programs to many small to mid-sized engineering teams to gain a reputation and potential paying clients.