AI Engineering Studio · Annecy — Geneva

AI agents that actually work for your business.

Nexelab designs, orchestrates and supervises custom AI agent systems: admin task automation, document processing, outreach, support — connected to your existing tools.

3multi-agent systems in production
10+years of software engineering
100%tested on our own companies

What we do

Creation, orchestration and management of AI agents

A single agent is a toy. A system of orchestrated agents, connected to your data and supervised over time, becomes a reliable collaborator.

Custom agent creation

Agents designed for your real processes, not generic chatbots.

  • Document and email processing
  • Business assistants (accounting, HR, quotes)
  • Connection to your tools (CRM, POS, ERP)

Multi-agent orchestration

Multiple specialized agents cooperating under a conductor, with clear rules.

  • LangGraph architectures / workflows
  • Routing, memory and guardrails
  • Human in the loop at key steps

Management & supervision

An agent in production is measured, evaluated and improved. We handle the run.

  • Decision observability and traceability
  • Continuous quality evaluation
  • Maintenance, costs and scaling

Method

From use case to production system

An iterative approach: we prove value on a reduced scope before industrializing.

01 — Scoping

Use case audit

We identify high-volume repetitive tasks and quantify the expected gain.

02 — Prototype

Pilot agent in 2–4 weeks

A first functional agent, tested on your real data, validated by your teams.

03 — Orchestration

System integration

Tool connections, multi-agent coordination, guardrails and human validation.

04 — Run

Continuous supervision

Dashboards, quality evaluation, cost optimization and evolutions.

Real cases

Systems already in production

Multi-agent assistant

ELIAS

Multi-agent personal assistant connected to WhatsApp: captures tasks from team conversations, classifies, schedules and follows up.

LangGraphFastAPIWhatsApp
Augmented research

Cytoscope

Multi-agent biomedical research system: literature analysis, citation validation and AlphaFold structural predictions.

Multi-agentsAlphaFoldRAG
Automated outreach

Enerbat

Automated outreach platform for energy renovation: targeting, generation and postal dispatch driven by agents.

AutomationData pipelineGenerative AI

Technical foundation

AI on the surface, 10 years of engineering underneath

A reliable agent rests on solid software architecture: robust APIs, mastered cloud, security and clean data.

Agent engineering

Claude / LLMsLangGraphMCPRAGEvaluation & observability

Platform & integrations

Python · FastAPINode.jsVue / NuxtFlutterAWS · GCPDocker · CI/CD

About

Vincent Vergès, founder

Full-stack developer for over 10 years and multi-sector entrepreneur (restaurants, building services, e-commerce), I build AI agents to solve problems I face myself daily. Based between Annecy and Geneva, I work directly, no middleman.

What task should your agents handle first?

30 minutes to identify together your most profitable use cases. No commitment, no jargon.