Inass Soukarieh

– Industrial MVPs & Production ML (L'Oreal) – Built and maintained end-to-end machine learning workflows, including data ingestion, preprocessing, modeling, validation, and deployment – Developed production-ready clustering and representation learning pipelines (PAM, PCA, MFA, deep embedding-based methods) for large-scale datasets – Created and maintained internal Python and R packages with automated testing, documentation, and CI/CD to ensure reproducibility and reliability – Collaborated with cross-functional teams to interpret model outputs, ensure transparency, and support data-driven decision-making – Standardized development environments using Docker Devcontainers and Git-based workflows for consistent execution across local and cloud environments – Implemented CI/CD pipelines using GitHub Actions and Google Cloud Build for automated testing, artifact management, and deployment – Deployed data processing and ML services on Google Cloud Platform (Cloud Run, Compute Engine, Cloud Storage, BigQuery) – Optimized large-scale data processing pipelines for performance, traceability, and robustness in regulated and production environments – Generative AI & Agentic Systems (PoCs) – Designed multi-agent AI systems using AutoGen, integrating planning, tool execution, iterative self-reflection, and coordinated agent collaboration – Developed agentic RAG pipelines using Llamaindex, enabling multi-document reasoning, synthesis, and summarization over large unstructured datasets

– Integrated machine learning models with multi-scale cardiac electrophysiology simulations for early drug safety assessment – Developed Deep Learning models under strong physical and domain constraints, emphasizing robustness, interpretability, and scientific validity – Designed deep learning models to predict cardiac action potentials under compound exposure, accelerating cardiotoxicity screening – Implemented Physics-Informed Neural Networks (PINNs) and advanced variants (HyperPINNs, meta-learning approaches) – Built scalable training and experimentation pipelines on AWS (SageMaker, EC2, S3), including dataset versioning and automated evaluation – Collaborated with computational chemistry and systems biology teams to integrate predictive models into compound design workflows

– Conducted research on statistical modeling and asymptotic behavior of stochastic processes, contributing to the development of new analytical tools for high-dimensional data analysis – Worked on extending limit theorems for U-empirical processes, with applications to robust estimation and inference in modern data science contexts – Taught and supported graduate-level coursework in applied data analysis using R, guiding students through real-world data workflows, exploratory analysis, modeling, and interpretation – Provided teaching assistance for undergraduate courses in real analysis and linear algebra, helping students build a strong foundation for advanced quantitative reasoning – Mentored students in statistical computing and analytical best practices, promoting reproducibility and rigorous model evaluation

– Conducted research on statistical learning methods and stochastic process modeling for complex and high-dimensional datasets – Developed theoretical and computational extensions of empirical process theory and U-statistics with relevance to modern data science applications – Designed statistical estimators and optimization strategies for regression, ranking, and non-parametric discrimination tasks under dependence structures (functional data, random fields, Markovian systems) – Applied statistical and machine learning techniques to real-world domains, including healthcare analytics and predictive modeling – Contributed to methodological development supporting robust inference and model performance in non-stationary and high-dimensional settings

– Conducted research on statistical modeling and inference for dependent data with applications to real-world stochastic systems – Developed and evaluated Bootstrap-based methods for Markov chain models to improve estimation accuracy and uncertainty quantification – Implemented parameter estimation pipelines, including transition matrix estimation and validation, using Python and statistical computing tools

– Developed computational models to study population dynamics and evolutionary processes using probabilistic and statistical frameworks. – Implemented simulation and inference algorithms in Python, including Markov Chain Monte Carlo (MCMC), random walks, Hidden Markov Models, and birth-death process modeling – Applied Bayesian inference (using BayesAss) to estimate migration rates and population structure from multilocus genotype data – Translated complex genetic modeling concepts into accessible formats for broader scientific communication and collaboration