Bastien Delaly

Technical Focus Areas: • OCR: Adapting current OCR techniques for complex documents analyses • Semantic search: Using LLM to keep context aware semantic search and remove hallucinations • LangChain & LangGraph: Building sophisticated AI agent workflows for complex reasoning and multi-step analysis • Neo4j & GraphRAG: Implementing knowledge graph architectures for enhanced contextual understanding • Full-Stack AI Applications: React/TypeScript frontend with FastAPI backend, production-ready deployment • AI Pipeline Architecture: Scalable processing systems with modern orchestration tools Key Learning Outcomes: → Mastered cutting-edge AI frameworks in production environment → Advanced from traditional ML to Graph-based AI reasoning systems → Full-stack AI product development from concept to deployment Expanded Technical Stack: • AI/ML: LangChain, LangGraph, GraphRAG, Advanced RAG architectures • Infrastructure: Neo4j, Docker, modern CI/CD, cloud deployment • Development: Full-stack ownership, React, TypeScript, FastAPI Keywords: LangChain, LangGraph, GraphRAG, Neo4j, AI Agents, FastAPI, Auth0, Full-Stack Development, Product Strategy

Key Responsibilities: • Designed and deployed end-to-end ML pipelines using BigQuery, Cloud Run, and Docker • Served as the technical liaison between Data, UA, and Creative Studio teams • Developed AI systems to support user acquisition and creative performance analytics Core Objectives: • Reduced LTV forecast latency to enable faster UA bidding and ASO optimization • Enhanced creative analytics using vision AI and vector search for asset tagging and duplication detection • Surfaced early signals of competitor virality • Automated trend scouting through multi-agent LLM chains Notable Projects: • Fast LTV Forecast Engine • Vision-based creative tagging and similarity search • Generative AI pipeline for social trend detection • Early-alert dashboard for creative virality Achievements: → New LTV model fully accepted by UA team → Vision-AI workflow scaled across creative teams → Launched internal tooling to detect and visualize emerging trends Keywords: Survival Analysis · LTV Forecasting · CLIP Embeddings · Vector Search · Vision AI · UA Analytics · Creative Intelligence · Generative AI · Trend Detection · Agentic Workflows Stack: Python, SQL, BigQuery, Cloud Run, Docker, Looker, Streamlit, GitLab, OpenAI & Gemini APIs, CrewAI, FFMPEG

Company Overview: Kovalee empowers mobile app creators. We use science and data to turn any app into a worldwide champion. We partner with app owners via the publishing model, and we bring their apps to the top charts thanks to our unique in-house tech and expertise. I witnessed Kovalee's transformation from a startup with just 5 full-time employees to a post-Series A scale-up boasting a dynamic team of 35 professionals. Key Responsibilities: - Built the entire data environment from scratch. - Led the data team. - Built a comprehensive data strategy for new data products, laying the foundation for continuous improvement over the next three years, including recruitment. - Sharing trustable data and insights. Core Objectives: - Accurate LTV forecasting of users a day post-installation for User Acquisition and ASO optimization. - Determining the winning variant in AB tests. - Enhancing data visualization techniques for clearer insights. - KPI monitoring and management. - Serving as the primary source for data-driven insights following my analyses and recommendations. - Architecting the entire data pipeline for optimized data flow. Notable Projects: - Developed forecasts using advanced statistical methods like Weibull extrapolation. - Leveraged machine learning techniques, specifically Survival Analysis, for precise forecasting. - Implemented ETL/ELT processes to ensure data integrity and availability. - Churn and Conversion Rate Analysis. - User behavior analysis. Achievements: -> Successfully achieved forecast accuracy exceeding 90%, even a year into predictions. -> Cracked SKAdNetwork. Keywords: Machine Learning, Lifetime Value Predictions, Weibull Extrapolation, Survival Analysis, AB Testing, Churn Analysis, Conversion Rate Optimization, SKAdNetwork, fROAS, CPI, and Optimization. Technical Proficiency: Languages & Tools: Python, SQL, VSCode Platforms & Services: Docker, Serverless, GitHub, AWS Suite (S3, SageMaker, Lambda, Glue, Athena), Snowflake

Context and objectives • SaaS company helping hoteliers maximize their revenue. • Start new innovating data-driven projects. • Integrate new data-driven tools in existing products. Keywords • Data • Data science • Data strategy • Data product management & development • Machine Learning • Natural Language Processing • Management of interns • Client facing Technical environment • Python • SQL • Google Cloud Platform • Git/GitLab • Tableau, Data Studio, Looker Python packages • Numpy • Pandas • Scikit-learn • XGBoost • Flask Projects As a data scientist, my projects' goals are to help hoteliers maximize their revenues, understand their market, ease the solution's integration, and reduce redundant and time-consuming onboarding and client service team tasks. Achievements • Built an application for restitution of the results of a classification algorithm. • Built automated processes for time series prediction. • Building clusters of hotels for market analysis with natural language processing. • Building a recommendation system. • Building tools for automated mappings. Personal assessment • Management • Self learning • Confidence

Context and objectives • IBM Almaden Research Center San Jose California • Electronic nose • Predictive algorithm development Keywords • Data analytics • Data science • Machine Learning • Deep Learning • Artificial Neural Networks • Convolutional Neural Networks (CNN) • Recurrent Neural Networks (RNN) • Long Short-Term Memory (LSTM) Technical environment • Python • Keras (Package machine learning sur python) • Numpy • Pandas • Scikit-learn • Linux Achievements • Task #1 : Data manipulation • Task #2 : Algorithm development - Subtask #2.1 : Data understanding - Subtask #2.2 : Choice of the best machine learning algorithm - Subtask #2.3 : Algorithm optimisation • Task #3 : Results analysis • Task #4 : Synthesis and presentation to high level managers Results • 99% accuracy for the predictive analysis Personal assessment • Python and Machine Learning skills development • Autonomy • Perseverance

Context and objectives • Master thesis • IBM Almaden Research Center San Jose California • Development of a sensor hub to detect home activity for aging-in-place Keywords • Data analytics • Data science • Classification Technical environment • Python • Raspberry Pi • Sensors (Motion, Temperature, Humidity, Pressure) • Numpy • Pandas • Scikit-learn • Linux • SQL Achievements • Task #1 : Project initiative - Subtask #1.1 : State of the art - Subtask #1.2 : Budget - Subtask #1.3 : Proposal to high level managers • Task #2 : Prototype - Subtask #2.1 : Proposal of a prototype - Subtask #2.2 : Manufacturing • Task #3 : Data acquisition • Task #4 : Data manipulation • Task #5 : Data analysis • Task #6 : Master thesis writting Results • Working prototype • Data acquisition environment not good enough • Difficulty to take full advantage of the results Personal assessment • Python skills development • Synthesis and writting skills development • Oral and wirtting english skills development

Nasim Annabi's Laboratory - http://www.northeastern.edu/annabi/ Professor Nasim Annabi’s laboratory is one of the numerous laboratories inside Northeastern University. This establishment is a private research University, founded in 1898 in Boston. It is categorized as a Doctoral University with Highest Research Activity, like 115 other universities through the United States. Its location, in Boston, makes Northeastern even more prestigious thanks to the neighborhood and the inevitable competition with Harvard University and the Massachusetts Institute of Technology. The principal interests of the laboratory are: - Biomaterials design - Tissue engineering - Nano- and micro-engineering technologies The main purpose of its research is to mimic as well as possible the native tissues present in the body, in order to do regenerative medicine like cardiac tissue regeneration or wound healing. My project was to engineer a cardiac patch to regenerate the heart capacities after a myocardial infarction. - Led a research effort to develop hydrogels for cardiac tissue regeneration. - Developed hydrogels using aptamers, growth factors and nanoparticles. - Performed mechanical and in-vitro and in-vivo testing.

Completed the process of machining and assembling of an electromagnet as workman in a factory