Model Factory
Custom continued pretraining, distillation, and model merging. Your premium differentiator for domain-adapted open models — Llama, Qwen, Mistral — delivered as a repeatable pipeline.
Open Model FactoryService provider playbook / Open weights
Built for a services firm selling outcomes, not GPUs. Each category below is a separate page with full delivery detail — what you sell, how you deliver it, and how you package it. Search to filter, click to dive deep.
Custom continued pretraining, distillation, and model merging. Your premium differentiator for domain-adapted open models — Llama, Qwen, Mistral — delivered as a repeatable pipeline.
Open Model FactoryManaged SFT, LoRA, QLoRA, DPO with eval-gated promotion, plus dataset engineering. Prove the tuned model beats base before it ships.
Open Fine-TuningOpenAI-compatible endpoints on vLLM / SGLang, sovereign VPC / on-prem / air-gapped deployments, and cost-optimized serving with routing and quantization.
Open InferenceBehavioral, topic, and cost drift monitoring, silent-swap detection for upstream weight changes, and eval-gated CI that blocks regressions.
Open Drift & EvalsSelf-hosted agent stacks with tool use that never leaks data, plus RAG-as-a-service on customer infra — vector DB, embeddings, retrieval evals.
Open Agents & RAGLicense and compliance audits for model provenance, plus closed-to-open migration programs with $/correct proof and drop-in URL swaps.
Open GovernanceEach page follows the same structure: customer problem and value, delivery mechanics, and commercial packaging with acceptance criteria. Use the left navigation to jump between subdetails. Toggle Theme for light mode in workshops.
Keyboard: press / equivalent by clicking the search box to filter cards. All pages work offline with no external dependencies — just open index.html in a browser.