Building machine learning systems that feel sharp, useful, and production-minded.
Available for ambitious ML engineering work
What I bring
Applied ML depth with strong engineering signal.
I am strongest when the work sits at the intersection of ambiguity, technical difficulty, and practical constraints. I turn vague, high-stakes problems into credible baselines, rigorous evaluation, and systems that still make sense outside the notebook.
- Structured experimentation under noisy, imperfect real-world conditions
- Judgment about the right level of model and system complexity
- Hands-on execution across modeling, software engineering, and integration
- Comfort in scientific, industrial, and product-facing environments
About
I like work that rewards judgment, speed of understanding, and strong follow-through.
My profile is shaped by applied machine learning, industrial software, computer vision, and scientific computing. That mix lets me stay practical without being shallow.
Approach
I like hard problems
I am drawn to projects where the data is imperfect, the requirements evolve, and success depends on engineering judgment as much as model choice.
Range
Broad without losing depth
My work spans retrosynthesis modeling, industrial computer vision, forecasting, VR simulation, biological data, and high-performance scientific computing.
Mindset
I think like an owner
I care about the full chain: framing, evaluation, implementation, failure modes, and whether the result is genuinely useful in practice.
Capabilities
Operating range
Tools and habits I can combine when a problem needs more than one kind of engineering.
Problem Framing
Evaluation Design
PyTorch
LLMs
Computer Vision
Forecasting
FastAPI
Docker
Linux
Python
C++
C#
OpenMP / MPI and CUDA
Unity
Real-Time Systems
Performance Optimization
Model Robustness
System Integration
Experience
Built in environments where the technical challenge was real and the work had to hold up.
Applied Machine Learning Engineer / Dassault Systèmes, Paris
Worked on retrosynthesis prediction under dataset shift using customer reaction data inside BIOVIA. The work focused on benchmarking, adaptation strategy evaluation, and defining deployment-relevant evaluation gates rather than only training a stronger model.
Software Engineer, VR & Computer Vision / Siemens, Munich
Built industrial VR applications and integrated vision-based tracking with real-time data pipelines. The work demanded product thinking, performance awareness, and strong execution inside larger industrial workflows.
Scientific Visualization Engineer / Internship at XLIM Lab
Developed high-performance molecular visualization methods in C++ and OpenGL, strengthening the low-level performance and numerical thinking that still helps me on complex technical systems today.
Education
Deep technical foundations behind the applied work.
Technical University of Munich
M.Sc. Computational Science and Engineering
Rigorous training across applied mathematics, computer science, and engineering with machine learning electives including Deep Learning, Computer Vision, HPC, parallel programming, and scientific computing.
Master's thesis: Scalable Kernel Matrix Inversion using Hierarchical Low-Rank Approximations, implemented for Gaussian processes, kernel ridge regression, and Kernel PCA, achieving a 20-30% speed-up.
Ecole Centrale de Nantes
Diplôme d'ingénieur
Equivalent to Bachelor + Master of Science in Engineering
Broad-based multidisciplinary engineering training through a highly selective program covering mathematics, computer science, industrial engineering, finance, and business foundations.
Final-year specialization in real-time computer graphics, VR/AR, computer vision, and industrial software engineering.
Results
Competition and project references that reinforce technical range and execution quality.
Kaggle Silver medal level
Image Matching Challenge 2025
Medal-level research-complexity computer vision pipeline work combining local features, dense descriptors, matching, and geometric verification with strong end-to-end engineering signal.
Kaggle Bronze medal
Stanford RNA 3D Folding
Bronze medal on a difficult scientific ML problem, showing fast adaptation in a deep biological structure domain.
Zindi Gold medal(TOP 5)
Hydropower Forecasting
Top 5 among 1,000+ participants on a forecasting problem with real operational relevance and noisy inputs.
Reference Project
Nexar Dashcam Crash Prediction
Sequence modeling and early-event prediction from video, combining perception with temporal reasoning and system-level thinking.
Reference Project
Scalable Kernel Inversion
Scientific computing work rooted in mathematical structure, computational efficiency, and implementation detail rather than surface-level ML experimentation.
Reference Project
Amini Cocoa Contamination
Applied computer vision under mobile-friendly constraints, which is useful evidence of practical engineering judgment and not just leaderboard optimization.
Selected Work
Selected work that shows applied range, competitive strength, and real engineering depth.
Applied ML Engineer at Dassault Systemes, with standout work in computer vision and scientific ML, including a medal-level Image Matching Challenge 2025 pipeline that reflects both research taste and production-minded execution.
Zindi / Gold
Repo ->
Hydropower Climate Optimisation
Ranked Top 5 out of 1,000+ participants in a forecasting challenge tied to hydropower optimization. Improved performance through feature engineering, validation design, and disciplined iteration under noisy conditions.
XGBoostForecastingFeature Engineering
Kaggle / Silver Medal level
Repo ->
Image Matching Challenge 2025
Built a medal-level image matching pipeline using learned dense features, keypoint detection, matching, and geometric verification. It is one of the clearest examples of full-pipeline vision engineering, balancing model choice, retrieval quality, and geometric robustness.
DINOv2ALIKEDLightGlueGeometric Verification
Zindi / Silver medal
Challenge ->
Amini Cocoa Contamination Challenge
Ranked 26th out of 300 in a computer vision challenge focused on cocoa leaf disease, with an emphasis on mobile-friendly deployment constraints.
YOLOComputer VisionMobile Constraints
Kaggle / Computer vision
Repo ->
Nexar Dashcam Crash Prediction
Built early collision prediction models from dashcam video using visual backbones and temporal modeling. This project reflects the kind of perception problem where timing and system design both matter. Very nice and complete project.
EfficientNetLSTM / GRU3D CNN
Kaggle / Bronze
Repo ->
Stanford RNA 3D Folding 2025
Ranked in the top 10% on a biologically complex structure prediction challenge, showing quick learning in a deep scientific domain with strong empirical execution.
Biological DataStructure PredictionKaggle
Scientific Computing
Repo ->
Scalable Kernel Inversion
Thesis-linked work on hierarchical low-rank approximations for kernel methods, combining mathematical structure with computational efficiency and implementation detail.
HPCKernel MethodsPerformance
What These Projects Show
More than isolated notebooks
Across these projects, the consistent pattern is not one niche benchmark. It is the ability to learn quickly, design credible evaluation, choose practical approaches, and execute across very different technical contexts.
Engineering Signal
Broad enough for junior growth, strong enough for senior ownership
My strongest value is being useful in hard environments: contributing hands-on as an engineer, but also reasoning clearly about failure modes, tradeoffs, and what makes a system deployable.
Contact
Looking for ML engineering work where the problems are genuinely difficult and the bar is high.
I am especially interested in roles with challenging data, demanding systems, high ownership, and room to solve problems that do not already come with a recipe.