CNC Collision Detection AI Engine

I need an AI engine that can open my CAM project files together with their 3D part, fixture and machine models, scan every programmed move, and flag any tool collision risks before a single chip is cut. The focus is solely on collisions with the machine, the fixture or the part; if the tool could crash, I want to know exactly where, when and why. The workflow I picture is straightforward: drop in the CAM file, let the AI analyse the toolpaths and geometry, then receive a clear, human-readable report plus a visual overlay that highlights the risky zones. A simple pass/fail headline is not enough—radius offsets, spindle positions and axis limits must all be considered so I can adjust the programme with confidence. Deliverables • A working collision-detection engine (library, plug-in, or stand-alone app) that accepts mainstream CAM formats and their associated 3D models. • A concise report output (JSON or CSV) and an optional 3D viewer/overlay showing collision points. • A sample dataset with at least three parts, each containing a confirmed collision, to demonstrate the engine’s accuracy and repeatability. • Brief setup notes or installer so I can reproduce your results on my workstation. Acceptance criteria • Detects every collision in the supplied test set and produces zero false positives on the provided clean paths. • Analysis completes within a practical timeframe for typical mould-making jobs (minutes, not hours). • Runs locally without requiring a live machine connection and without uploading proprietary geometry to third-party servers. If the core works well, I’m open to extending the system later to warn about air-cuts or sub-optimal feeds, but for this first phase collision detection is the entire mission.