Conversational AI
& Multi-Agent Systems
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.
What we build
| 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) |
Production track record
Tangible proof, not promises.
Multi-agent recommendation platform
Tourisfair — orchestrates specialised agents (sentiment, recommendation, planning). Python, LLMs, RAG, CustomGPT. Production, EU+US multi-region.
RAG pipeline + search engine
Traveler reviews → LLM-based recommendation. Python, embeddings, PostgreSQL. Production — semantic search at scale.
Messenger chatbot agent
Conversational travel planning. NodeJS, NLP, LLM orchestration. 1000+ user sessions in production.
Sentiment analysis engine
"Feeling-based" search on user corpus. Python, ML, NLP. Production — real-time analysis.
SaaS platform designed to scale
Wizipet — embedded AI features. Python, TypeScript (NestJS), Docker. Architecture designed to scale to 100,000+ users — a design target, not reached in production.
Why me (not a web agency, not an LLM reseller)
"The LLM hallucinates on my regulated data"
Chain-IT implements strict RAG with guardrails, cited sources, and human validation on high-risk answers. No black box.
"I don't want to be locked into OpenAI"
Agnostic stack — I work with all LLMs and design for vendor switching with zero exit cost.
"How do I put this in production in a regulated environment?"
25+ years of experience in GAMP5, FDA, 21 CFR Part 11 → traceability, audit trail, validation built into the architecture.
"The agent runs wild without supervision"
Human-in-the-loop designed in from day one. The agent proposes, the human validates. No autonomy on risky decisions.
My delivery approach
From scoping to production in 4 structured phases.
- Week 1 — Scoping
Business + IT interviews. Data review. Target architecture.
Volume, quality, sensitivity. SI constraints. Functional scope.
Functional spec + architecture validated - Weeks 2–4 — Structured POC
First agent or pipeline development. Real user testing.
No fake demos. Real users, real cases, real feedback. Fast iterations.
Functional POC with real users - Weeks 5–8 — Production
SI integration (SSO, ERP, API). Security hardening. Monitoring.
Logs, alerts, backup, non-regression. Technical documentation.
System in production + runbook - Week 9 — Handover
Internal team training. Documentation. Post-go-live support.
2-week post-production support. Autonomous team.
Team trained and autonomous
Use cases by sector
DORA compliance RAG agent
Instant answers for legal/compliance teams on contracts and regulations.
Time reduction: 70%Claims triage chatbot
Automatic qualification, routing, key information extraction.
Cost reduction: 30–50%Predictive maintenance agent
Sensor log analysis, intervention recommendations before failure.
Unplanned downtime: –20–40%Supply chain assistant
Route optimisation, anomaly detection, corrective proposals.
Logistics costs: –10–20%Augmented internal knowledge base
Employee Q&A (HR, IT, procedures).
Support tickets: –40%Indicative budget
Document Q&A / RAG
Simple, 1 document source
€15k–€20k
Conversational chatbot
Business, 1 use case
€20k–€30k
Multi-agent system
2–3 orchestrated agents
€25k–€40k
Critical regulated system
Integrated governance, audit trail, validation
€30k–€50k
Next steps
- Qualification call (30 min) — You describe your use case, I tell you if it's a good fit
- Technical proposal — Architecture, timeline, budget, guarantees
- Scoping — We define specifications together (1 week)
Have an AI use case?
Chain-IT has designed and deployed multi-agent systems in production. Let's build yours.