Skip to main content

AI Engineering

Applied AI that earns its keep

Claude, GPT, Gemini, Mistral, open-source LLMs. RAG on enterprise data, ML Ops, anomaly detection, conversational self-service.

Talk to an architect
Access Copilot

Identigy Access Copilot — an AI Copilot for IGA

Admin AI chat, an MCP server, RAG over your policies, on the LLM you already run. Vendor-neutral — on top of Evolveum midPoint, SailPoint or Oracle.

Open the Copilot page
AI · Services Six engagement types

What we deliver

From single-purpose assistants to full ML Ops platforms. The AI × IDM intersection is our differentiated track.

Service catalogue

  • Enterprise LLM assistants

    Claude, GPT-4, Gemini, Mistral or open-source LLMs (Llama 3, Qwen, DeepSeek). On-prem when sovereignty matters.

  • Retrieval-Augmented Generation (RAG)

    RAG over enterprise data with pgvector, Qdrant, Weaviate. Chunking, evaluation harnesses, citation quality.

  • Multi-agent orchestration

    Agent frameworks (LangGraph, custom). Tool use, MCP servers, audit trail, prompt-injection defence.

  • Custom ML

    Anomaly detection (cross with IDM access analytics), classification, NER, OCR, recommendation engines.

  • ML Ops and data engineering

    Airflow / Dagster pipelines, feature stores, drift detection, containerised inference (Docker / K8s).

  • AI × IDM intersection

    Access-pattern anomaly detection, onboarding role recommendation, conversational self-service workflows.

Model stack

  • Claude · GPT-4 · Gemini · Mistral
  • Open-source (Llama 3 / Qwen / DeepSeek)
  • On-prem for data-residency cases
  • pgvector · Qdrant · Weaviate

Discovery

Bring a problem statement — we'll come back in 48 hours with a fixed-fee proof-of-value proposal.

Talk to an engineer →
AI × IDM Cross-pillar

Where AI meets IDM

The differentiated track — AI applied to the access management problem set, designed and operated by engineers who run IGA every day.

Use cases

Anomaly detection on access patterns

Detect risky combinations of permissions before recertification — not after.

Auto-recommendation of roles

Onboarding-time role bundles based on department, peers and historical patterns.

Auto-classification of access criticality

Tag entitlements by sensitivity to drive recertification cadence and SoD weighting.

Conversational self-service

Natural-language access requests that route to the right SailPoint or midPoint workflow.

Why this works

IGA data — entitlements, role hierarchies, peer groups — is structured. AI applied to that data gets concrete, auditable outcomes rather than hallucinated suggestions.

See IDM practice

The AI × IDM track lives inside our wider Identity Governance programme. Many AI engagements start as an extension of an ongoing IGA programme.

See IDM →