Charbel Dandjinou

AI R&D Engineer - MLOps Tech Lead / Principal GenAI Developer

Lyon, France

French (Native) • English (C1/Advanced) • Gungbe (Native) • Japanese (Beginner) • Fon (Native) • Yoruba (Advanced) • Ewe (Advanced)

Professional Summary

Deep tech builder with expertise in domain-specific generative AI, autonomous infrastructure deployment, and strong theoretical background. Active AI evangelist, educator, team leader, and research contributor. Self-learner reading ~2 AI papers daily. Innovative thinker who solves problems in original ways and invents new methods to tackle complex challenges.

Professional Experience

Currently holding two concurrent leadership positions:

Generative AI Programme Lead / Principal Developer

Oct 2024 - Present

Exxact Robotics / EXEL Industries Group

  • Lead 3 RAG programs (industrial, HR, legal); judge-guided RAG reaching near-perfect legal-QA relevance on 3k+ queries (statistically significant vs baseline)
  • Multimodal RAG: ColPali-style late interaction for layout/tables/images + text; hybrid dense+BM25; adaptive chunking; LangChain/LangGraph orchestration
  • Training: QLora + RLHF (DPO/GRPO) for domain Q&A; embedding fine-tuning (contrastive, hard negatives); cross-encoder reranking
  • Serving: vLLM/Ollama, KV cache, PagedAttention, quantization (AWQ/GPTQ/GGUF) → 2-3× throughput on A100
  • Reliability: MRR/MMR, ablations, significance testing; sovereign deployments (no external APIs)
  • Delivered 15+ internal/external GenAI talks and workshops to 200+ attendees

MLOps Tech Lead

Aug 2024 - Present

Exxact Robotics / EXEL Industries Group

  • Own company-wide AI infra: Kubernetes (on-premise), GitOps (ArgoCD), observability (Prometheus/Grafana/Loki), MLflow
  • Standardize packaging, datasets/artifacts, reproducible deployments; mentor DevOps and AI apprentice
  • Define SLOs for latency/GPU; implement data/feature drift monitoring in production
  • Ensure AI security, compliance, and data sovereignty for enterprise deployments

Solo MLOps Engineer

Mar 2024 - Jul 2024

Exxact Robotics / EXEL Industries Group

  • Created and executed company-wide MLOps roadmap from scratch
  • Built full MLOps stack: MLflow server, dataset versioning, artifact store, benchmark tools
  • Implemented Infrastructure as Code (IaC) using Terraform with GitOps practices (ArgoCD)
  • Set up on-premise GPU Kubernetes cluster with complete observability stack
  • Delivered entire AI infrastructure single-handedly with 100% cost control vs cloud

AI Research & MLOps Engineer

Sep 2023 - Feb 2024

Exxact Robotics / EXEL Industries Group

  • Trained segmentation models for crop and weed detection
  • Built Optuna-based tool for augmentation strategy search using Bayesian optimization
  • Researched conformal prediction for active learning
  • Delivered production-ready reproducible pipeline

AI Research Engineer + MLOps Onboarding

Jun 2023 - Sep 2023

Exxact Robotics / EXEL Industries Group

  • Transitioned to MLOps while maintaining research responsibilities
  • Set up MLflow experiment tracking infrastructure
  • Deployed first MLOps PoC using cloud infrastructure
  • Technologies: ZenML, MLflow, Kubernetes (GKE), MinIO, GCP bucket

AI Research Engineer (Apprenticeship)

Oct 2022 - Jun 2023

Exxact Robotics / EXEL Industries Group

  • Designed PowerBALD-based active learning pipeline for semantic segmentation with WUR
  • Developed RL-based data augmentation search using Bayesian optimization
  • Created custom augmentation tool based on Optuna for hyperparameter tuning
  • Achievement: Reduced annotation costs by 30%
  • Co-authored paper presented at ICCV CVPPA 2023

Freelance AI Developer - Computer Vision Consultant

2020 - 2021

Remote

  • Designed and implemented object detection systems for diverse applications
  • Retail theft detection: Real-time monitoring and alert systems
  • Agricultural anomaly detection: Crop disease and pest identification
  • Delivered end-to-end computer vision solutions to multiple clients
  • Helped businesses automate surveillance and quality control processes

AI Research Intern - Computer Vision

Jun 2019 - Sep 2019

Laboratoire PRISME, Orléans

  • Master 2 research internship on deep learning for embedded systems
  • Achieved +70% accuracy improvement over classical computer vision methods
  • Optimized neural networks for deployment on low-power Linux boards
  • Built Qt-based demonstration GUI for industry partners (LIFAT + LENZI)
  • Research Stack: C++, Qt, TensorFlow, OpenCV, model optimization

Full-Stack Developer (Intern)

May 2018 - Nov 2018

Fonddeuzz

  • Improved web portal UX
  • Prototyped deep-learning recommendation engine
  • Delivered first working recommender demo to founders
  • Technologies: HTML/CSS/JS, Node.js, React, TensorFlow-JS

Auxiliaire de vacances

Feb 2020

BNP Paribas, Pantin

  • Admin & archiving on ALIS platform

Technical Skills

Model Training & Optimization

Advanced PyTorch (DDP/FSDP), mixed precision (AMP/BF16/FP16), custom training loops

Augmentation strategy & HPO (Optuna), regularization, early-stopping, TTA

Data/feature drift detection & monitoring

Evaluation: IoU, mAP, PR, MRR/MMR, ablations, significance testing

Deep Learning & AI/ML

PyTorch, TensorFlow, Hugging Face Transformers

Active Learning: Baal, PowerBALD

Optimization: Optuna, Bayesian optimization

Computer Vision: OpenCV, semantic segmentation, object detection

Methods: Conformal prediction, RL-based data augmentation

Generative AI & LLM Operations

Fine-tuning: QLora/LoRA, RLHF (DPO/GRPO) for domain Q&A; embedding fine-tuning (contrastive, hard negatives)

Retrieval: Dense bi-encoders (FAISS-GPU, ChromaDB), cross-encoder reranking, layout-aware pipelines (ColPali)

RAG Methods: Hybrid dense+BM25, adaptive chunking, query reformulation; LangChain/LangGraph orchestration

LLM Serving: vLLM, Ollama, SGLang; KV cache, PagedAttention; quantization (AWQ/GPTQ/GGUF)

Agents: Smol Agent, MCP (Model Context Protocol)

MLOps & Infrastructure

Orchestration: Kubernetes (GKE & on-premise), Docker, Helm

IaC: Terraform, Ansible

CI/CD: ArgoCD (GitOps), GitHub Actions

ML Platforms: MLflow, ZenML

Storage: MinIO, GCS, Ceph, NFS

Web Development & Design

Frontend: React, Vue.js, Next.js, Tailwind CSS, Canvas API

Backend: Node.js, Express, RESTful APIs, GraphQL

UI/UX: Complete design process, Figma, Adobe Creative Suite

Creative Coding: Three.js, D3.js, real-time visualizations

Visual AI & Creative Tech

Diffusion Models: Stable Diffusion, FLUX, Midjourney

Generative Models: Autoregressive models (GPT-style), VAE, GAN architectures

Fine-tuning: LoRA, DreamBooth, Textual Inversion, Hypernetworks

Control: ControlNet, IP-Adapter, T2I-Adapter, Pose guidance

Workflows: ComfyUI advanced pipelines, A1111, InvokeAI

Applications: Product design, concept art, fashion tech

Monitoring & Observability

Stack: Grafana, Prometheus, Loki

Custom: Structured logging, metrics collection, GPU monitoring

Embodied AI & Robotics

NeuroGraph: Long-horizon memory (vector retrieval + dynamic knowledge graphs with temporal decay)

VLA Models: VLM/VLA integration, memory-conditioned control

Agentic Physical AI: LLM-robot tool use (MCP), safety gates

Frameworks: LeRobot (Hugging Face), ACT, Diffusion Policy

Platforms: ROS2 Humble, Isaac Sim; sim-to-real transfer

Research Contributions & Publications

Co-author: "PowerBALD: A Hybrid Active-Learning Approach for Semantic Segmentation" – ICCV CVPPA 2023 (extended abstract + poster)
Internal Research:
  • "Novel Hybrid RAG Architecture for Domain-Specific Question Answering" (2024) - Achieved significant MRR increase
  • "Conformal Prediction for Active Learning in Agricultural AI" (2024)
  • "Cloud-to-On-Premise Migration Strategies for AI Infrastructure" (2024)
Invited Reviewer: CVPPA Workshop @ ECCV 2024 (reviewed 3-4 papers)

Research Metrics

• 730+ AI papers analyzed (~2/day continuous learning)

• 5+ novel algorithms developed and deployed to production

• 30-99% improvement metrics across all research projects

• 4+ researchers/engineers mentored

Key Achievements

Near-perfect relevance on internal legal-QA with judge-guided RAG (3k+ queries; statistically significant vs baseline)

≈30% cost reduction labeling cost via PowerBALD active learning (CVPPA @ ICCV 2023)

>70% accuracy gain replacing classical CV with YOLOv3 (2019, PRISME)

On-prem GPU cluster ML platform built from scratch; improved latency and cost vs cloud

Invited Speaker GOSIM 2025; ICCV workshop 2023 and arXiv 2024; reviewer (ECCV 2024 workshop)

200+ attendees at 15+ GenAI presentations as AI evangelist

Founded IROKO international NGO bridging AI innovation between Europe and Africa

Created CD_IA website with real-time fractal animations and dual-theme design

Featured Talk

Leadership & Community Initiatives

IROKO Association

Founder & President • Feb 2025 - Present

  • Founded international NGO supporting emerging-market talent worldwide (primary focus: Africa)
  • Mission: Democratizing AI knowledge and fostering innovation in underserved communities
  • Organize AI seminars and "AI Afterwork" networking events connecting global talent
  • Build bridges between developed and emerging AI ecosystems
  • Develop partnerships with universities and tech hubs internationally
  • Impact: Creating pathways for underrepresented talent in global AI research

Deep Takka (IROKO Research Lab)

Head of AI Research • May 2025 - Present

  • Launched research lab as IROKO's technical innovation arm
  • Leading Project Genesis: foundational research initiatives
  • Focus: Culturally-aware AI, low-resource languages, edge AI optimization
  • Building research team and establishing international collaborations
  • Status: Multiple research projects in development phase

Academic Mentoring & Supervision

Current Supervision

Master's Apprentice - GenAI Research Assistant (2024-Present)

  • Direct supervisor and mentor for Master 2 AI student
  • Research: Multimodal RAG systems, evaluation frameworks
  • Impact: Accelerated GenAI project delivery by 40%

Past Supervision

Robotics Learning Research Intern (2024)

  • Led hiring process and supervised research
  • Topic: NVIDIA Eureka framework for robotic task learning
  • Successfully integrated findings into production pipeline

Neural Architecture Search Research Intern (2023)

  • Co-supervised research on automated model optimization
  • Focus: NAS for resource-constrained edge deployment

Side Projects & Community

Creative & Technical Projects

  • Personal Website (CD_IA): Designed and developed dual-theme portfolio with real-time fractal visualizations (Julia sets, Sierpinski triangles, glitch patterns) using vanilla JavaScript and Canvas API
  • Generative AI Artist: Creating product designs and concept art using custom-trained diffusion models
  • Robot Learning: Building hobby projects with ACT, VLM/VLA models, and Diffusion Policy
  • LeRobot Contributor: Active experimentation with Hugging Face's robot learning framework

Community Engagement

  • Hackathon Mentor: Mentored participants at LeRobot Hackathon
  • AI Fairness: Developing inclusion datasets for minority cultural items
  • Object Recognition: Fashion-tech applications for global textiles and patterns
  • Education: AI education initiatives in emerging countries
  • Open Source: Contributing to LeRobot, Hugging Face ecosystem, and community projects

Education

Master 2 Artificial Intelligence & Management

IA School, Paris • 2022-2023 •

Master 1/2 Computer Science (DataScale)

Université Paris-Saclay • 2019-2021 •

Master Systems & Networks

IFRI-UAC, Benin • 2017-2019 • C

Additional Information

Security Clearance: Experience with sovereign AI systems (no external APIs)

Compliance: AI security and compliance management

Languages: Fluent in technical English for paper reviews and conferences

Availability: Full-time position

Location Preference: Open to discussion