Use cases

From research labs to production APIs, from cloud-hosted to edge devices — Polymodels adapts to wherever you need lightweight, composable model specialization.

Domain specialist agents

Train experts on legal documents, HVAC manuals, medical literature, or product catalogs. Route customer queries to the matching expert on a prompt-by-prompt basis — all running on the same model, serving multiple users concurrently.

Classic promptsAlpacaFunction (Gemma)

Function calling & tool use

Use the FunctionGemma prompt format to train experts that reliably emit structured tool-call JSON. Keep the base model available for open-ended generation; route only tool-call prompts through the expert. Or use multiple experts for different tool sets or completely different domains.

FunctionGemmaStructured outputTool use

Comparative evaluation

Run the same prompt through multiple experts in one batch. Different training datasets, different prompt types, different hyperparameters — results arrive side by side. No separate deployments needed.

Batch inferenceGate indicesA/B testing

Research & experimentation

Study targeted layer-level interventions. Share a single GPU across a team of researchers, each with their own expert slots. Compare training configurations without restarting the model or affecting colleagues.

Multi-userLayer expansionTraining queue