Architecture, methodology and impact for each engagement.
Detailed architecture and complete technical analysis.
Three coordinated Vertex AI pipelines that scrape product reviews, harmonise product identities via embeddings and deliver multilingual sentiment signals into BigQuery.
View full case studyWeekly ML pipeline that detects under-indexed SKUs and delivers 30 priority visit recommendations to each sales rep every Monday, fully automated on Azure.
View full case study
LLM pipeline that reads invoice emails, classifies each document via few-shot learning and updates Eloficash in real time, with a human review dashboard.
View full case study
LLMOps Lead on Secure GPT, an internal generative AI conversational assistant. Ingestion, inference and LLM evaluation pipelines. Autonomous multi-agent orchestration for claims handling with Agent-to-Agent (A2A) communication.
View full case study
Development of innovative analytical solutions in Python/C++ based on clustering, optimisation and probabilistic models to significantly increase 5G throughput.
View full case study
Agentic RAG system for automated translation editing and revision. Document ingestion pipeline, semantic retrieval and augmented generation to ensure terminological consistency and translation quality.
View full case study
From a Spotify track, the system fetches the lyrics and audio data, has an LLM analyse the themes, finds related films (soundtracks, web search), checks their streaming availability via JustWatch, then ranks the results by embedding similarity between the lyrics and film synopses.
View full case study
Web platform for Institut Viavoice: AI-assisted coding of open-ended questions (code plan generated in parallel batches, coding, validation, Excel export), extraction of result tables from Excel, chart suggestions and generation of PowerPoint decks with native charts.
View full case study
CAWI survey programming tool: a Word questionnaire is imported by AI (Claude or GPT) and converted into a structured survey (routing, rotations, termination rules). Screen editor, quotas, respondent links, response tracking, analysis and standalone HTML export.
View full case study
Data engineering for Aware, a Brussels retail big data start-up: collecting, structuring and making retail data reliable for analytics.
View full case study
Conversational assistant deployed in production for TheFork: users describe what they want in natural language and get relevant restaurants based on context, real-time availability and preferences, with LLM monitoring.
View full case study
Prototype built for PwC: the stand-up as a conversation, with sprint dashboard, blockers ranked by priority and history. Interactive mock-up with sample data.
View full case study
Internal project: an agent graph collects AI news (blogs, arXiv, Reddit, Substack), clusters articles by topic, spots trends and writes the newsletter.
View full case studyBook a call: we scope your need and propose a concrete plan of action.