An engineering company, not “AI adoption”
Anchor Inference builds LLM agents, RAG search and self-hosted inference so they can be deployed in a closed customer perimeter, verified and maintained.
How we handle data
Data lives where you say: your cloud, on-prem or a closed perimeter. We do not move the corpus to our side and do not use external APIs to process your documents.
Models run locally through vLLM. The choice of open models is agreed per task and hardware: we measure quality and cost instead of promising “the best model”.
Code ownership is fixed in the contract. We do not promise certifications we do not hold.
Why “anchor”
On a nautical chart an anchor marks a place where a ship can lie safely. We anchor an AI system inside your perimeter: it works, it can be verified, and it does not carry data out into the open sea of foreign clouds.
Hence the chart on the home page: the query's route, four soundings and an anchorage. See the chart.
Founder
Aleksandr Tikhonov, ML/AI engineer. Focus: agent architectures, RAG and semantic search, self-hosted inference, evals.
Company status
Anchor Inference LLP, Astana, Kazakhstan. The company is being registered. Legal details will appear here once it is entered in the registry.
Until then work is done under a contractor agreement; the form of cooperation is discussed on the first call.
How we contract
A contract with a company, not hiring a specialist on staff. Contracting a Kazakhstan LLP does not require a foreign-labor permit. This is not legal advice: check with your counsel.