Mbogol Touye Achille

– Designed, developed, and deployed multi-agent Al systems (LLMs + tools) enabling real-time interaction with the Monstock cloud platform – Built multimodal Al agents (OCR + LLMs) to automatically populate the platform and execute real business operations – Implemented automation pipelines to : – Analyze supplier documents (quotes, purchase orders, invoices) – Automatically generate purchase and customer orders – Detect, alert, and correct invoice inconsistencies – Developed Al-driven workflows executing end-to-end business actions, including: – Inventory monitoring and stock adjustment – Stock transfers and operational analytics – Order creation from multichannel inputs (email, voice, chat) – Logistics process triggering – Automated notifications (email, SMS) – Designed and deployed multimodal Agentic RAG architectures for internal knowledge and operational use cases – Integrated structured/unstructured data sources for contextual reasoning and decision-making – Designed and deployed scalable forecasting models (demand, inventory, replenishment) using PySpark on AWS EMR – Enabled: – Accurate demand prediction – Inventory optimization – Automated replenishment workflows – Maintained and optimized AWS EMR clusters – Deployed Al systems on AWS (EMR / EC2-based clusters) – Implemented CI/CD pipelines (GitLab) for automated deployment, versioning, and monitoring of Al services – Mentored junior data scientists and supported their technical growth – Collaborated with clients to translate business needs into scalable Al solutions

– Teaching Big Data fundamentals (Hadoop, Spark) to students, focusing on scalable data processing – Delivering courses on Deep Learning (CNN, LSTM, xLSTM, VAE, GAN, Transformers) and Large Language Models, including text generation and fine-tuning – Training students on MLOps practices with MLflow, DVC, and CI/CD (GitHub Actions, Docker, Heroku), along with model monitoring using Evidently

Mission client (Airbus) : – Implemented and deployed multimodal Agentic RAG, to process technical raw PDF – Evaluate the performance of LLM : Similarity-Evaluator, Retrieval-Evaluator, QA-Evaluator, Groundedness-Evaluator, Coherence-Evaluator – Build CI/CD Pipeline with Azure Pipeline to automate deployment and testing of GenAl components – Tracing to view performance and Monitoring dashboards to supervise GenAl agents using Azure monitor – Tools & Technologies : Vs code, Python, Azure Al Search, Azure Al computer vision/ OpenClip, Azure OpenAl, Langchain, Langgraph, Azure Blob storage, Azure Pipeline, Docker, Azure VM, Azure Monitor. Mission client (Renault) : – Implemented and deployed Al Agent to analyze customer reviews for automotive manufacturers : enabling sentiment classification (to categorize reviews as positive, negative, or neutral and identify key topics, themes, and specific mentions related to the automotive products or services) – Evaluate the performance of LLM: Precision / Recall / F1-score /Accuracy/ Similarity-Evaluator, Retrieval-Evaluator, QA-Evaluator – Build CI/CD Pipeline with Azure Pipeline to automate deployment and testing of GenAl components – Tracing to view performance and Monitoring dashboards to supervise GenAl agents using Azure monitor in Azure – Tools & Technologies : Vs code, Python, Azure Al Search, Azure Al computer vision, Azure OpenAl (GPT-4, Whisper), Langchain/langgraph, Azure Blob storage, Azure Pipeline, Azure Storage, Docker, Azure ACR and Azure VM, Azure KeyVault Monitor. Mission client (Dassault Systèmes) – Reconstruction old Document via OCR – Implemented and deployed multimodal Agentic RAG On Premise, to process technical old raw PDF Mission Interne (Big data Forecasting of air quality) : – Collected and cleaned structured datasets, applied feature engineering. – Compared various Al models (LSTM, xLSTM, sLSTM, mLSTM, CNN-LSTM, Transformers) – Automated the CI/CD workflows (Gitlab), including unit testing, deployment triggers, and rollback mechanism – Integrated monitoring tools for drift detection, and anomaly alerts, ensuring production-grade reliability via Evidently Al

– Develop and implement high-level educational content and tools in the field of Al, particularly in deep learning by using GPU Accelerator from IDRIS and Gricad Cluster. – Implement examples, use cases, and practical work in collaboration with research teams and teachers. – Contribute to the monitoring of participants in training sessions directly implemented by FIDLE. – Collaborate directly with multidisciplinary educational and research teams. Organize discussions with various project stakeholders, including teachers and learners. Computer Vision Engineer R&D Collaboration with Industry Framatome for many tasks : – Semantic Segmentation of Metallography image for Quantification & Classification microstructure of weld metal using UNet model (Fine-tuning) – Predicting mechanical properties of Weld metal from chemical composition using Neural Network model. – Co-supervise Master Student at CNRS Institut Neel in Quantum Chemistry : – We propose to explore machine learning methodology to predict the magnetic properties of metal-organic-frameworks (MOF) by using Graph Neural Network (GNN). – Provide guidance and support to master's students during their internships. Ensure the alignment of their projects with educational and research objectives. – Training Program By EFELIA-MIAI, Support, By the University of Grenoble-Alpes Computer Science Practical Work for L3 Physics students at University of Grenoble : – Teach the basics of programming Python language (Python fundamentals, Object-oriented programming – File handling and exceptions, modules and packages, code optimization, developing projects, Scikit-Learn, numpy, pandas) University of Limoges (remote teaching) Master's Student Acyson (mathematics and application) : – Basic knowledge of deep learning neural networks – Learn the computational tools to use and implement deep learning algorithms (CNN, LSTM, AE, VAE, Transformers) & LLM

– Modelling, FEMM Simulation, Experimental, Data analysis, Signal processing. – Characterizations of ferroelectrics, ferromagnetics & multiferroics materials. – Work collaboratively with Kapteos SAS Supplier of measuring instruments. Valorization and Transfer of Research Results. – Present the results at Limerick Sensors Conference. Write reports and scientific papers for all results. – Instructor of Instrumentation & Computer Science L3 students: – Pratical Works on Circuits & components electronics, LTSpice, Probability, descriptive & inferential statistical, Python & scikit-learn, numpy, pandas.