Developed the theoretical framework and implemented MIIC_search&score in R, a novel search-and-score algorithm for causal discovery from nonlinear and non-Gaussian data with latent variables.
Built a large-scale benchmarking framework evaluating five causal discovery methods across 960+ graph configurations on HPC clusters (SLURM, PBS). MIIC_search&score outperformed all tested methods on nonlinear and non-Gaussian benchmarks, while improving precision over the MIIC baseline by 18% on continuous and 27% on categorical data. This work led to a first-author publication at ICML 2025.
Designed and implemented MIIC-to-DAG in R, an algorithm converting partially oriented graphs into DAGs while preserving key causal structures, later integrated into the MIIC-SDG synthetic health data generation framework published in npj Digital Medicine.
Contributed to the open-source MIIC R/C++ package and extended the existing MIIC web interface into MIIC-Display, enabling interactive visualization of locally generated networks without uploading sensitive data.
Co-mentored two master’s research interns, providing day-to-day scientific guidance.
Data Analysis Consultant
Sorbonne Université
Developed reproducible Python workflows for cleaning, analyzing, and visualizing student survey data.
Applied statistical testing, dimensionality reduction, and clustering to evaluate hybrid and project-based teaching programs and investigate student learning behaviors.
Presented analytical findings to an interdisciplinary team and contributed to a co-authored publication at QPES 2025.
Teaching Assistant
Sorbonne Université
Provided 80+ hours of practical teaching support in Python and C to first-year undergraduate students, helping with debugging, programming concepts, and practical assignment grading.
Delivered master’s-level practicals in biological networks and systems biology using R, and helped supervise projects using real-world biological and healthcare datasets, providing technical guidance and contributing to project assessment.
Bioinformatics Research Intern (M.Sc. Thesis)
I2BC & Université Paris-Saclay
Developed ksub, an RNA-seq pipeline implemented in Python and C++ for detecting tumor-associated transcripts in lung adenocarcinoma using complementary reference-based and reference-free approaches.
Processed 1.4 TB of RNA-seq data (17 billion reads, 213 tumor and normal samples) and built large-scale indices containing up to 494 million k-mers.
Designed a multi-stage filtering strategy combining sequence assembly, large-scale indexing, and statistical modeling, reducing the candidate transcript space by 99.85%.
Characterized the resulting transcript landscape, identifying known lung adenocarcinoma alterations alongside non-coding RNAs, immunoglobulins, repetitive sequences, and viral transcripts.
Ranked 1st out of 10 students. Master’s thesis: Towards a Complete Molecular Portrait of an Individual Tumor Using RNA Alone Research conducted at I2BC under the supervision of Prof. Daniel Gautheret.