Skip to main content

Technology

How Chervik works

An educational overview of the AI and infrastructure technologies that power the Chervik platform.

01

AI Models & Multi-Model Routing

Chervik uses a model-agnostic routing layer that selects the optimal LLM for each task — balancing cost, latency, and capability. Supports OpenAI, Anthropic, Mistral, Llama, and custom fine-tuned models. Models run locally or via private API endpoints.

OpenAIAnthropicMistralLlama 3Custom Models
02

Retrieval-Augmented Generation (RAG)

RAG grounds AI responses in your actual enterprise data, eliminating hallucinations. Documents are chunked, embedded, and stored in a vector database. At query time, the most relevant chunks are retrieved and injected into the LLM context window alongside the user's question.

ChunkingEmbeddingsVector SearchContext WindowGrounding
03

Hybrid Search

Pure semantic search misses exact matches. Pure keyword search misses intent. Chervik combines both using a reciprocal rank fusion algorithm — delivering the precision of BM25 with the semantic understanding of vector search.

BM25Vector SearchReciprocal Rank FusionRe-ranking
04

Security Architecture

Security is not an afterthought. All data is encrypted at rest and in transit. Role-based access control ensures users only see content they're authorised to access. Every AI query is logged for audit purposes. Supports deployment in air-gapped environments.

Encryption at RestTLSRBACAudit LogsAir-Gap Support
05

Cloud-Native Architecture

Built on Kubernetes for portability across AWS, Azure, GCP, and on-premise. Infrastructure is defined as code using Terraform. Horizontal scaling, health checks, and zero-downtime deployments are built in.

KubernetesTerraformAWSAzureGCPHelm
06

Data Sovereignty

Your data never leaves your infrastructure. Chervik is designed to run entirely within your cloud account or on-premise data centre. No telemetry, no data sharing, no vendor lock-in on your knowledge assets.

On-PremisePrivate CloudNo Data SharingGDPR Ready

Ready to deploy AI that stays in your control?

Talk to our team about your enterprise AI requirements. No commitment required.