Home / Products
Three product lines, one fine-tuning pipeline.
Self-hostable enterprise RAG, a language program for under-served languages, and bespoke fine-tuning for your data. All built on the same engineering bench — and all designed so you keep the weights.
Tabotec Atlas
Self-hosted retrieval-augmented generation platform. Document upload, vector search, cited answers. Plugs into Anthropic, OpenAI, Gemini, or your local llama-server.
- FastAPI + Postgres + pgvector
- Multi-workspace with auth and roles
- Prometheus metrics and OTel tracing
Tabotec Languages
A department of Tabotec building speech and language AI for under-served languages — starting in the Horn of Africa with Tigvoice (ASR/TTS) and an instruction-tuned Tigrigna LLM.
- Tigvoice — Tigrigna speech recognition & synthesis
- Tabotec Tigrigna — instruction-tuned LLM (Geez script)
- Roadmap: Amharic, Afaan Oromoo, Somali, and beyond
Models built for you
Bring your data, your domain, your success criteria. We curate, train, evaluate, and deliver a model you can host or call as an API. No black box, no surprise bills.
- ASR, TTS, LLM, embeddings, multimodal
- Eval suites built around your tasks
- You own the resulting weights
Need help, not a product?
If what you actually need is an AI team to help you architect, evaluate, or ship — that's our consulting practice. Same engineers, same opinions, scoped to your reality.
- Discovery sprints and architecture reviews
- Build engagements (4–12 weeks)
- Fractional AI lead engagements
Not sure which one?
Tell us what you're trying to do and we'll point at the right starting point — or tell you honestly that we're not the right fit yet.