Post-training an LLM for math reasoning, knowledge, multilinguality and safety with LoRA, SFT, DPO and RLVR — then merging the specialists with TIES-Merging.
Data Science & Financial Engineering
Cyrine Akrout
Master's student at the École Polytechnique Fédérale de Lausanne (EPFL), passionate about machine learning, language models and quantitative finance — and about turning rigorous math into things that work on real data.
Who I am
I'm Cyrine, a Data Science & Financial Engineering master's student at EPFL, holding a Master's Excellence Fellowship.
I studied Communication Systems at EPFL and spent a year at the University of Toronto, where I explored software engineering, security and financial engineering. Today I work at the crossroads of machine learning and quantitative finance: post-training language models, building deep networks for vision, pricing derivatives and designing systematic investment strategies.
I also work as a student assistant at EPFL, helping students in finance, probability, algorithms and signal processing. I love learning in teams and I'm looking for opportunities where I can keep growing while building things that matter.
French · nativeEnglish · C1Arabic · B2
Education
2025 – Present
MSc in Data Science & Financial Engineering
École Polytechnique Fédérale de Lausanne – EPFL
Machine Learning · Modern NLP · Applied Data Analysis · Large-scale Data Science · Stochastic Calculus · Derivatives · Quantitative Risk Management
2024 – 2025
Applied Science & Engineering (exchange year)
University of Toronto
Software Engineering · Computer Security · Random Processes · Financial Engineering · Artificial Intelligence
2022 – 2024
BSc in Communication Systems GPA 5.89/6
École Polytechnique Fédérale de Lausanne – EPFL
2021 – 2022
Preparatory Year (CMS) GPA 5.8/6
École Polytechnique Fédérale de Lausanne – EPFL
Skills
Python
PyTorch
TensorFlow
NumPy
Pandas
scikit-learn
Jupyter
Java
Scala
C
SQL
React
TypeScript
Firebase
Git
LaTeX
LLM post-training (LoRA · SFT · DPO · RLHF)Semantic segmentationObject detectionStochastic calculusDerivative pricingPortfolio constructionWeights & Biases
Experience
Feb 2026 – Present
Student Assistant
EPFL
Principles of Finance · Probability & Statistics · Algorithms · Signal Processing
Jun 2025
Machine Learning Intern
Exypnos
Built an object-recognition pipeline for infrared images with lightweight deep learning models (YOLOv9, RT-DETR-Lite): preprocessing, training, inference and benchmarking on real IR data.
Sep 2024 – Dec 2024
Software Engineer — BalanceAI
Dr. Jang's research lab with OISE, University of Toronto
In a team of six, developed gamification features for an AI literacy-assessment platform for K-12 students, using React, TypeScript, Next.js and Firebase. balanceai.ca ↗
Projects
Data Science & Machine Learning
PyTorch pipeline on ~5,000 aerial images comparing U-Net, U-Net++, D-LinkNet and DeepLabV3. 3rd place on AIcrowd — F1 0.922, accuracy 0.958.
Classification on the BRFSS health survey with robust preprocessing and regularized logistic regression.
MLP, CNN and Vision Transformer compared on Fashion-MNIST, with PCA for dimensionality reduction.
Linear regression, logistic regression and k-NN implemented and evaluated on the Stanford Dogs dataset.
Infrared Object Detection
Internship · Exypnos
Lightweight detectors (YOLOv9, RT-DETR-Lite) for object recognition in infrared images, benchmarked on real IR datasets.
Quantitative Finance
Valuing dividend futures and options under Black-Scholes and affine stochastic dividend models — Feynman–Kac, Riccati ODEs, calibrated to market data.
Carry, momentum, dollar, risk-parity and mean-variance strategies on six G10 currencies — carry turns out to be the strongest predictor of FX returns.
Beyond the code
Singing & karaoke
Music is my favourite way to switch off. I love to sing, and karaoke nights with friends are some of my best moments: give me a mic and a good song, and I'm happy.
Music
Always listening to something new.
Oriental dance
Rhythm and expression.
Thriller novels
The more plot twists, the better.
Contacts
Have an opportunity, a question about a project, or just want to say hi? Send me a message here and I'll get back to you.