Freelancer profile translated to English.
Back to original languageAbout Pujols
Applied AI — autonomous agents, LLM, hybrid RAG, vector databases, tool calling, MCP protocols, multi-agent orchestration, computer vision, and edge AI — I design systems connected to real data, software, and processes. An entrepreneur for 19 years, I connect strategy, technical architecture, product, and execution to transform AI opportunities into gains in speed, margin, or capacity.
At Apple, I developed SOFIA, an AI-augmented mental health system: personalized journeys, clinical data structuring, conversational interfaces, and actionable support protocols. With Troublemaker AI, I integrate embedded models, sensors, computer vision, and decision loops into B2C physical products, creating devices that adapt to their use rather than just adding a screen. With Avalanche, I worked on a carbon capture approach relying on autonomous robots, computer vision, and AI to analyze terrain, guide operations, and significantly increase grassland productivity.
My work involves identifying high-value bottlenecks and then building the right stack: data ingestion and cleaning, RAG with citations and access control, specialized or frontier models, model routing, evaluations, guardrails, memory, agents connected to CRMs, ERPs, documents, and internal tools. I favor robust architectures: human-in-the-loop when necessary, observability, traceability, security, controlled costs, and sovereign or on-premise deployment.
Today, with SOFIA AI SYSTEMS and Troublemaker AI, I support companies and leaders in creating business agents, intelligent SaaS platforms, connected products, and decision-making systems. My goal: reduce lead times, eliminate repetitive tasks, accelerate sales, and make AI an operational asset with a demonstrable ROI.
French
Native or bilingual
English
Native or bilingual
German
Fluent
Can work on-site
Paris 02 Bourse (up to 50km)
Experience
- AppleSenior Product ManagerTECHDecember 2024 - February 2026 (1 year and 2 months)Paris, FranceBuilding S.O.F.I.A., an AI-augmented mental health system developed as an internal venture at Apple, with the support of TMC.S.O.F.I.A. explores how machine learning, conversational AI and multimodal behavioural signals can make evidence-informed mental health support more accessible, personalised and useful in everyday life.The system translates evidence-based approaches from Third-Wave CBT—including ACT, MCT and DBT—into adaptive digital journeys, guided exercises, reflective tools and practical interventions designed to help users build emotional regulation, cognitive flexibility and healthier behavioural patterns over time.The product combines LLM-based interaction with structured therapeutic protocols, behavioural science, UX research and privacy-by-design principles. Rather than offering generic wellbeing content, it aims to adapt support to an individual’s context, needs, progress and engagement patterns—while preserving clear safety boundaries and keeping clinicians central when higher-intensity care is needed.The ambition is to move beyond traditional wellness apps and create a scalable, consumer-facing mental health system: one that helps people identify difficulties earlier, develop practical psychological skills and access more relevant support before distress becomes acute.
- SOFIA AI SYSTEMSCTO/CAIO/Senior AI EngineerTECHMay 2024 - Today (2 years and 2 months)Paris 02 Bourse, FranceSOFIA AI SYSTEMS designs and deploys AI systems that create measurable ROI: less wasted time, more production capacity, faster decisions, and better revenue capture.We are no longer in the era of demos, decorative chatbots, or POCs that never go into production. Every mission starts with a concrete business process — sales, support, operations, documentation, management, finance — and an indicator to improve: processing time, cost per case, response rate, conversion rate, volume processed, errors avoided, or revenue generated.I design RAG, LLMs, AI agents, and automations truly integrated into your environment: CRM, ERP, Microsoft 365, Google Workspace, Slack, Teams, Notion, SQL databases, APIs, and business tools. Systems can search your internal knowledge, generate sourced answers, qualify requests, enrich data, prepare actions, or execute workflows, with human validation where necessary.Production architecture: Python, APIs, Docker/Docker Compose, vector databases, relational databases, n8n, MCP, webhooks, CI/CD, monitoring, access control, European cloud, hybrid, or on-premise.SOFIA AI SYSTEMS is not tied to any vendor. “Not a dog in the fight”: I select the best assembly of models, open-source or proprietary, based on your expected quality, costs, data, security, and independence. No unnecessary layers or oversized models: I “trim down” AI to achieve maximum efficiency with minimum complexity and cost.Methodology: ROI scoping, data and tools audit, targeted prototype, secure deployment, results instrumentation, continuous improvement. The goal is not to sell you AI. It is to deliver a reliable system that your teams use, understand, and that pays for itself.
- Troublemaker AICEOTECHJanuary 2022 - Today (4 years and 6 months)Paris 07 Palais-Bourbon, FranceTROUBLEMAKER creates products and services where AI, software, and physical objects form a desirable and minimalist experience.I handle everything from idea to production: product strategy, AI integrated into connected devices and B2B equipment, SaaS platforms, prototyping, and UX/UI. My approach combines Human-Centered Design, technical rigor, and an obsession with real-world use: every interaction should reduce friction, improve decisions, or create measurable value.AI is never an afterthought. It's applied where it simplifies the user experience and makes the product more performant: agents, RAG, sensor data, conversational interfaces, automation, and embedded systems.Troublemaker aims for more responsible AI: appropriately sized models, optimized computing, durable hardware, reduced unnecessary processes, and lifecycle-conscious design. Consumer or B2B: less impact, more intelligence, an experience that users want to engage with.
Recommendations
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- AI Engineer - IDE - specialization AI (Double Master generalist and AI/DevOps)Imperial College London2019Spécialisation appliquée en intelligence artificielle, données et ingénierie logicielle, construite sur la méthodologie Innovation Design Engineering : transformer une technologie complexe en produit, système ou service réellement utilisable, désirable et économiquement viable. Le parcours couvre la conception de systèmes IA de bout en bout : analyse de besoins métier, architecture data, choix de modèles, prototypage, intégration aux outils existants, déploiement et mesure de la valeur créée. Compétences développées et appliquées : Architecture de solutions IA génératives : LLMs, modèles multimodaux, modèles open source et APIs de modèles frontier. RAG et knowledge systems : ingestion documentaire, OCR, nettoyage de données, chunking, embeddings, bases vectorielles, recherche sémantique, recherche hybride, reranking, citations et contrôle de la qualité des réponses. Agents IA et workflows agentiques : tool calling, orchestration, MCP, mémoire, règles métier, human-in-the-loop, supervision et journalisation des actions. Data engineering : structuration de données, pipelines ETL/ELT, APIs, webhooks, SQL, bases relationnelles, vector databases et synchronisation avec les outils métiers. Automatisation : intégration CRM, ERP, support, messagerie, documentation et outils collaboratifs via n8n, Make, Zapier ou développements sur mesure. Production engineering : Python, FastAPI, Docker, Docker Compose, services conteneurisés, CI/CD, gestion des secrets, logging, monitoring, contrôle des accès et déploiement cloud, hybride ou on-premise. Sécurité et souveraineté : GDPR, gestion des rôles, confidentialité, hébergement européen, architectures privées, limitation des permissions et protection contre les prompt injections. Human-Centered AI : UX/UI, interfaces conversationnelles, conception centrée utilisateur, tests d’usage, adoption des équipes et réduction de la complexité opérationnelle. Évaluation et ROI : benchmarks, jeux de tests, mesure de précision, coût par tâche, latence, temps économisé, réduction des erreurs, capacité opérationnelle et impact commercial.
- Master in Management — Grande École ProgramHEC Paris2016Formation généraliste de haut niveau en stratégie, finance, entrepreneuriat, transformation des organisations et management international. Le programme développe une compréhension complète de l’entreprise : de la construction d’un modèle économique à son exécution opérationnelle, en passant par la prise de décision, la gestion de projet, le marketing, la négociation et le pilotage de la performance. Cette formation m’a apporté une capacité à relier enjeux business, technologie et expérience utilisateur. Elle structure aujourd’hui mon approche chez SOFIA AI SYSTEMS : identifier un problème à forte valeur, prioriser les cas d’usage, construire une solution réalisable et mesurer son impact économique réel. Compétences mobilisées : stratégie d’entreprise, business models, analyse financière, gestion de projet, transformation digitale, innovation, conduite du changement, leadership, négociation, développement commercial et pilotage du ROI.