ENSEEIHT, Toulouse
Concurrent systems, networks, databases, computer architecture, formal methods. High-performance computing internship: Kubernetes, Linux, Bash, vLLM.
- Kubernetes
- Docker
- Linux
- Kyverno
- NGINX
- Bash
Polytechnique Montréal / ENSEEIHT
Sami Mostfa
I design ML models and their deployment to production.
Professional Master's student in Computer Engineering, looking for a Summer 2027 internship in Montréal (May 3 – August 31) in AI/ML Engineering, MLOps or Data Engineering.
A dual skill set
Between software engineering and AI, I work where ML projects usually hit reality: deployment.
Polytechnique Montréal
AI, ML, data mining, Bayesian statistics, NLP, continuous optimization.
Cumulative GPA of 3.83/4.0.
- PyTorch
- scikit-learn
- pandas
- XGBoost
- Julia
My profile: from model to production. Training a model, evaluating it rigorously, deploying it: that is the chain I am aiming for. At Axians, the solution I designed was put into production in 2026; I now want to take part in that chain end to end in a new professional opportunity.
My journey
-
AI/ML Engineering, MLOps or Data Engineering: leading a project end to end, from design to production.
-
LPA → LPAm → LPAm+ → VNS on 9 DIMACS graphs (up to 31,000+ vertices), 40 runs per method and a review of 5 papers.
-
Fine particulate forecasting across 7 stations from 17 years of data: 20+ features, temporal validation, Kaggle RMSE of 18.25.
-
Cumulative GPA of 3.83/4.0. AI, ML, data mining, Bayesian statistics, NLP, intelligent DevOps, cybersecurity.
-
Kubernetes integration with Open OnDemand (in production since 2026), Bash kubectl wrapper, Kyverno policies, vLLM LLM inference server.
-
Multi-output CNN (classification + bounding-box regression), IoU and F1 implemented from scratch in TensorFlow.
-
Pool supervision and emergency response: responsibility and decision-making in critical situations.
-
Software engineering, concurrent systems, databases, graph theory, computability, networks, formal methods.
-
Intensive mathematics, physics and computer science.
-
Mathematics and computer science: about 15 high-school students, 80 hours, alongside studies.
Projects
Three projects from my GitHub: from statistical modeling to graph optimization, plus an operations-oriented study (MLOps).
Graph Clustering by Modularity Maximization (VNS)
Implementation of a pipeline from the literature for community detection in graphs, through modularity maximization.
- LPA → LPAm → LPAm+ → VNS/VNDS: each step escapes the local optima of the previous one
- 9 benchmark graphs, from 34 to over 31,000 vertices, 40 runs per method and graph (seeds 1 to 40)
- Comparison with the literature, review of 5 scientific papers
- Julia chosen for near-compiled speed with Python-like readability
- Graph theory
- Metaheuristics
- Optimization
Air Quality Forecasting in Montréal
Forecast the daily concentration of fine particulate matter (PM) across 7 Montréal stations for all 365 days of 2025, i.e. 2,555 predictions, from 17 years of data.
- Air + weather data join, exploratory analysis, 20+ engineered features (rolling means, sine/cosine seasonality, temperature × month interactions)
- Missing values handled per variable; temporal split validation, never random
- 4 variants compared; selected model: linear regression (GLM.jl, 23 variables) with no feature built from past PM, to avoid error propagation
- RMSE of 18.25 on Kaggle; limitations documented: underestimation of wildfire smoke episodes
- Time series
- Linear regression
- Feature engineering
Disk Failure Prediction: an Operations-Centered Approach
Study of hard-drive failure prediction from an operations standpoint: the goal is not just a good score, but a prediction that is useful in time.
- Centralized, reproducible configuration: prediction horizons, minimum warning times, critical SMART attributes, cost ratios
- Schema audit of daily CSV files, conversion to compressed Parquet, queries with DuckDB
- Literature review: 5 papers on failure prediction and cost-sensitive learning
- Current phase: Q1 2025 exploration to frame the problem statement; extension to the full year planned
- Data engineering
- Cost-sensitive learning
- DuckDB / Parquet
Experience
06/2025 – 08/2025
HPC Engineering Intern
Solution put into production by Axians in 2026
- Designed an architecture integrating Kubernetes with the Open OnDemand HPC portal, working around the native adapter's limits (no Deployments)
- Bash
kubectlwrapper that intercepts the adapter's calls and makes job status display reliable - Kyverno policies: orphaned resource cleanup, quotas, access hardening (NFS, GPU)
- Deployed and configured an LLM (vLLM) inference server on the cluster
- Technical architecture document (Apache, NGINX, Passenger, Kubernetes adapter)
09/2020 – 06/2025
Private Tutor in Mathematics & Computer Science
~15 high-school students, ~80 hours. Communication, adaptability and autonomy over several years.
07/2024 – 09/2024
Lifeguard
Pool supervision and emergency response: responsibility and decision-making in critical situations.
Education
Polytechnique Montréal
AI, Machine Learning, Continuous Optimization, Data Mining, Data Visualization, Bayesian Statistics, NLP, Intelligent DevOps for Large-Scale Software Systems, Cybersecurity.
ENSEEIHT
Software engineering, concurrent systems, databases, graphs, computability & complexity, networks, formal methods.
Preparatory Classes (CPGE) MPI* / MP2I
Intensive mathematics, physics and computer science.
Skills
Programming languages
- Python
- Julia
- C
- Java
- TypeScript
- JavaScript
- OCaml
- Ada
- Bash
- HTML/CSS
- LaTeX
ML, AI & Deep Learning
- PyTorch
- TensorFlow
- Keras
- scikit-learn
- XGBoost
- GLM.jl
Agentic AI
- MCP
- OpenRouter
- Claude Code
- Prompt engineering
- Custom skills & tools
- vLLM
Data & visualization
- NumPy
- SciPy
- Pandas
- SQL
- Seaborn
- Matplotlib
- D3.js
- DataFrames.jl
DevOps & infrastructure
- Git / GitHub
- GitHub Actions
- Kubernetes
- Docker
- Kyverno
- NGINX
- Apache
- GCP
- Linux
Spoken languages
- French (fluent)
- English B2 (Cambridge Linguaskill)
- Spanish A2
- Arabic (basic)
Interests
Music
14 years of guitar, advanced music theory, 10 hours of lessons taught.
First aid & lifesaving
400 cumulative hours: French National Lifeguard Certificate (BNSSA, 8 months of training) and First Aid Team Responder Level 1 (PSE1, 40 hours).
Sports
Swimming & Weightlifting.