Skip to content
HANA PlatformHANARAD

AI stack

Our AI technology stack: models, frameworks and deployment

The models, frameworks and infrastructure we use to build chatbots, autonomous agents and knowledge assistants — integrated with the same secure web and backend stack as the rest of your software.

Discuss an AI projectExplore AI services

Technologies in our AI stack: OpenAI GPT, Claude, Gemini, Llama, DeepSeek, Mistral, LangChain, LlamaIndex, CrewAI, AutoGen, MCP, Ollama, Pinecone, Weaviate, Python, FastAPI, Node.js, Docker, Kubernetes, Redis.

The AI technology stack

Mainstream tools, chosen per use case

We are model-agnostic by design. Each category below is a toolbox we know deeply, so we can pick what fits your data, budget and privacy requirements.

  • AI models

    We choose per use case — accuracy, speed, cost and data residency — and design the application so the model can be swapped as better options appear. Open-weight models can run entirely on your own servers.

    • OpenAI GPT
    • Claude
    • Gemini
    • Llama· open-weight
    • DeepSeek· open-weight
    • Mistral· open-weight

  • Frameworks & protocols

    Orchestration for retrieval, multi-step agents and tool use, plus local model serving.

    • LangChain
    • LlamaIndex
    • CrewAI
    • AutoGen
    • MCP
    • Ollama

  • Vector databases

    Semantic search over your documents and records, the foundation of accurate retrieval-augmented generation.

    • Pinecone
    • Weaviate

  • AI backend

    Python services for model and data work, connected to our Node.js and NestJS application backend.

    • Python
    • FastAPI
    • Node.js

  • DevOps

    Containerised, reproducible deployments that scale with demand and queue long-running AI work.

    • Docker
    • Kubernetes
    • Redis

Integration

AI that plugs into our standard web and backend stack

AI is never a separate island. Every AI feature sits behind the same NestJS API, with the same validation, permissions, logging and tests as the rest of your software — so it is secure and maintainable by any developer in our pool.

  1. Step 01

    Your apps & systems

    Web app, mobile app, CRM, ERP or internal tools — the places your people already work.

  2. Step 02

    NestJS API · /v1

    Authentication, role and record-level permissions, input validation and rate limits on every AI request.

  3. Step 03

    AI service · Python + FastAPI

    Retrieval, prompts, agents and tool calls, with long-running work queued through Redis and BullMQ.

  4. Step 04

    Models & vector store

    Cloud model APIs or local models via Ollama, grounded in your data through a vector database.

Deployment options

Cloud, on-premise or hybrid — your choice

Where your AI runs is a business decision about privacy, cost and capability. We support all three and help you choose.

  • Cloud

    Hosted model APIs and managed infrastructure for the fastest start and access to the most capable models.

    Best for: Fast pilots, variable workloads and tasks that need frontier-model quality.

  • On-premise

    Open-weight models served on your own hardware or private cloud, so prompts and documents never leave your network.

    Best for: Strict confidentiality, data-residency rules and predictable high-volume usage.

  • Hybrid

    Sensitive data stays on local models while less sensitive or harder tasks route to cloud models, under one policy.

    Best for: Organisations balancing privacy, cost and capability across many use cases.

Running models privately? See our local LLM deployment service.

FAQ

AI technology stack: frequently asked questions

Have an AI use case in mind?

Book a free 30-minute consultation. We will assess feasibility, suggest the right models and deployment option, and outline a measurable pilot.

Get a Free 30-Minute Consultation
Call us WhatsApp

Keep exploring

Related pages