Jonathan Nguyen

• Lead the agentic architecture of the AI Slide Generator, turning a user prompt into a polished, brand-compliant structured deck, serving 130M+ corporate users. • Designed and built the slide generation evaluation pipeline, end to end and per component, and raise the generation quality score from 20% to 80% • Improved the MCP connection, RAG system, cost and latency of the generation pipeline and Office integration. • Technical stack: Azure, LLM, C, Typescript, React, RAG, LLMOps

A startup AI co-pilot for data science and machine learning. • Develop an AI-powered copilot using AI agents that streamline data science and ML workflows through automation, intelligent code assistance, and accelerated experimentation. • Design and implement a full stack architecture using Next.js, TypeScript, Python, Databases, and Google Cloud.

- Building a real-time deep learning solution that leverages advanced computer vision algorithms to detect and track fraud at self-checkout registers. - Driving business impact with a targeted 2-3% revenue increase for Carrefour through enhanced fraud detection capabilities. - Deploying edge computing solutions at each store for faster, more efficient processing and real-time insights. - Developed a package based on Vertex AI to help data scientists deploy their machine learning models into production more easily and quickly

- Build an Ad Network, an intermediary between advertisers, who are looking to place their ad in another app and publishers who has many apps running, that aims at generating 20% total revenue of company in 2 years. - Build and deploy machine learning models and recommendation systems at scale in Real Time Bidding for more than 100.000 impressions per second, over 40 million daily active users, and to compete performance with other Ad networks. - Build a bidding system that automates User Acquisition bidding on mobile Ad networks platforms for more than 200 games and a $100 million annual budget.

Saint Gobain Smart Pricing - Predict the price of sales promotion on specified items and specified clients Classification problem: Use decision trees models (xgboost, lightGBM, catboost) and analyze variables to classify whether a seller should forger the prices of some specified products Regression problem: Predict the price of sales promotion which can help a seller to fix the prices. Skills: python (scikit-learn, pipeline...), performing analysis with large datasets

The Banque de France rating is an assessment by the Banque de France of the company's ability to meet its financial commitments over a three-year horizon. It is made up of two components : a turnover rating and a credit rating. Mission: - In charge of the attribution of the "Banque de France" credit risk rating analyzing the capacity of a company to meet its commitments over a three-year period through accounting (balance sheet, cash flow statement) and financial data and qualitative analysis - Organised meetings with CEOs and CFOs of SMEs to discuss expected performance