The world's first nesting software that uses Large Language Models as direct optimisation algorithms. 12+ algorithms, a self-learning ML selector, and natural language explainability for every placement.
Every commercial nesting package runs the same algorithms. Magpie is the first to use LLMs as direct optimisation engines — with a self-learning ML layer that gets smarter with every job.
Magpie orchestrates a hierarchy — from millisecond ML predictions to deep LLM reasoning — to find the optimal strategy for every job.
A trained GAT model analyses every job in milliseconds. Each part becomes a node with geometric features — area, convexity, aspect ratio — connected by edges encoding part-to-part relationships. The output: a recommended algorithm and a confidence score.
When GAT confidence drops below threshold, or the job is too complex for a single algorithm, the LLM strategy advisor activates. It reasons about batching, algorithm selection, and parameter tuning — without generating coordinates.
The patented core: the LLM receives part geometries, sheet constraints, and manufacturing tolerances — then directly outputs placement coordinates and orientations. A validation pipeline catches overlaps and retries with targeted error feedback.
Four steps from file to factory floor. The complexity is ours, not yours.
Magpie orchestrates a pipeline of ML models, LLMs, and validation layers to find the best possible layout.
NFP, genetic, guillotine, B1 tournament, particle swarm, simulated annealing, and interlocking strategies — each tuned for different part geometries.
The ML model's confidence score determines whether to use fast deterministic algorithms or invoke the LLM for deeper reasoning. No wasted API calls.
Every LLM-generated layout passes through automated overlap detection, bounds checking, and constraint validation — with error-specific feedback and intelligent retry.
For high-stakes jobs, multiple LLMs independently evaluate layouts. A proposer, critic, and orchestrator debate strategy across providers — converging on the best result.
Every other nesting tool gives you a layout and asks you to trust it. Magpie tells you the reasoning behind every placement — in plain English.
Operators can verify the AI considered their constraints. Quality managers can audit decisions. The system builds trust because every choice is transparent.
Algorithms informed by hands-on manufacturing experience. Every constraint — grain direction, tool paths, minimum gaps, cutter tolerances — is handled because we've worked with them first-hand.
Every plan includes the full AI pipeline. Start small, scale when you're ready.
Book a live demo and see Magpie's AI pipeline optimise your real parts on a real sheet. Watch the LLM reason about your layout in real time.
Request a Demo →