Jean-Folly Kpodar

Context: Septeo is a software publisher by legal, notarial, and real estate branch. As part of improving and adding value to their software, the company decided to introduce generative AI into their solution to provide their clients with assistance in validating notarial deeds to alleviate the daily workload of notarial clerks.

My role on this project consisted of:

 Collection of needs from business teams for business use cases for data extraction from notarial documents

 Detailed functional/technical specification of the solution

 Choice of technical solutions on the market based on functional and non-functional needs

 Batching and piloting the implementation of the MVP

 Technical coordination of infrastructure teams for model deployment

• Océrisation Platform

• Fine Tuning and supervised transfer learning on notarial documents for classification

• Setting up prompts for data extraction

• Setting up APIs for document extraction in transactional mode

 Proposal of different architecture options

• RAG (Retrieval augmented Generation) + Vectorization/Prompt tuning to overcome LLM token issues for voluminous documents

• TAG (Table Augmented Generation) for tracking metrics and KPIs

• Agentic for batch extraction

 Implementation of an MVP to validate the solution (ReactJS and FastApi)

 Support for integration into products

 LLMOPS (langsmith) MLOPS(Mlflow)

Technical environment: Langchain, LangGraph, LangFlow, Langsmith, Qwen, Llama3, phi, Mistral, Azure OpenAi, Embeddings, Pyhton, OCR (tesseract, Azure Document intelligence), Groq, React, FastApi, CrewAI, CursorAI, Unstructured, GPT4o, TensorFlow, Scikit-learn

Context: Implementation of a Datalake for data analytics and Data science use cases for the company PROMAN

My role as Tech Lead Data consisted of:

• Deploying the SnowFlake solution on AWS

• Choosing and deploying ELT solutions for data ingestion into the Datalake

• Supporting juniors in the implementation of data pipelines Technical environment: SnowFlake, Fivetran, Informatica, Python/Snowpark, AWS, DB2, Sql Server, MySql, Terraform

As part of managing the obsolescence of their bigdata platform, URSSAF decided to migrate the Hortonworks bigdata platform to Cloudera. The project team consists of a PO, an internal Tech Lead, 2 technical architects, and 3 administrators. My role in this project was to provide dual expertise on these two platforms as a technical architect to secure the migration