I design, develop and deploy custom conversational AI and multi-agent systems — RAG pipelines, business chatbots, specialised agent orchestrators — for operational use cases in regulated environments.
| System type | Application example |
|---|---|
| Document Q&A / RAG | Agent answering compliance team questions on DORA/AI Act from your document corpus |
| Conversational chatbot | Claims triage assistant (insurance), technical support (industry), employee onboarding (HR) |
| Multi-agent orchestrator | Central orchestrator coordinating specialised agents: sentiment + recommendation + planning |
| Augmented search engine | Semantic search across your document base (contracts, procedures, audit reports) |
| CustomGPT | Private, specialised LLM version for your domain (clinical, regulatory, supply chain) |
Tangible proof, not promises.
Tourisfair — orchestrates specialised agents (sentiment, recommendation, planning). Python, LLMs, RAG, CustomGPT. Production, EU+US multi-region.
Traveler reviews → LLM-based recommendation. Python, embeddings, PostgreSQL. Production — semantic search at scale.
Conversational travel planning. NodeJS, NLP, LLM orchestration. 1000+ user sessions in production.
"Feeling-based" search on user corpus. Python, ML, NLP. Production — real-time analysis.
Wizipet — embedded AI features. Python, TypeScript (NestJS), Docker. 100,000+ users in production.
Chain-IT implements strict RAG with guardrails, cited sources, and human validation on high-risk answers. No black box.
Agnostic stack — I work with all LLMs and design for vendor switching with zero exit cost.
25+ years of experience in GAMP5, FDA, 21 CFR Part 11 → traceability, audit trail, validation built into the architecture.
Human-in-the-loop designed in from day one. The agent proposes, the human validates. No autonomy on risky decisions.
From scoping to production in 4 structured phases.
Volume, quality, sensitivity. SI constraints. Functional scope.
✅ Functional spec + architecture validatedNo fake demos. Real users, real cases, real feedback. Fast iterations.
✅ Functional POC with real usersLogs, alerts, backup, non-regression. Technical documentation.
✅ System in production + runbook2-week post-production support. Autonomous team.
✅ Team trained and autonomousInstant answers for legal/compliance teams on contracts and regulations.
📉 Time reduction: 70%Automatic qualification, routing, key information extraction.
📉 Cost reduction: 30–50%Sensor log analysis, intervention recommendations before failure.
📉 Unplanned downtime: –20–40%Route optimisation, anomaly detection, corrective proposals.
📉 Logistics costs: –10–20%Employee Q&A (HR, IT, procedures).
📉 Support tickets: –40%Simple, 1 document source
€15k–€20k
Business, 1 use case
€20k–€30k
2–3 orchestrated agents
€25k–€40k
Integrated governance, audit trail, validation
€30k–€50k
Chain-IT has designed and deployed multi-agent systems in production. Let's build yours.
