Description

AI/NLP engineer, 4+ years shipping ML systems from prototype to production. I lead R&D at a cybersecurity scale-up, designing and deploying LLM and deep learning pipelines end to end.

Strong in Python and PyTorch. Recent work: LLM applications (OpenAI API, LangChain, RAG, LoRA/PEFT), transformer architectures for time-series, and generative models for synthetic data — published at IEEE and IFIP conferences.

I deliver LLM pipelines, custom deep learning models, anomaly detection, time-series forecasting, and data extraction infrastructure. I do the research and ship the system. Let's talk.

Papers

CRAFT — Cross-Modal Anomaly Detection (IFIP SEC 2026): First-author work pairing temporal and textual branches to detect anomalies in cybersecurity telemetry; strong F1 across three industrial benchmarks.

Channel-Aware Embeddings for Time-Series Transformers (IEEE ICTAI 2025): First-author work cutting transformer parameters 83% with no accuracy loss in multivariate forecasting.

Ransomware detection: AI-driven detection of ransomware from system behavior telemetry, using deep learning and synthetic data generation.

Langues

Français
Capacité professionnelle limitée
Anglais
Bilingue ou natif
Hindi
Bilingue ou natif

Préférences en matière de lieu de travail

En télétravail uniquement

Travaille majoritairement à distance

SitInCloud
Head of R&D / AI Engineer (CIFRE PhD)
janvier 2024 - Aujourd'hui (2 ans et 6 mois)
Paris, France
• • Cybersecurity scale-up — OwlyShield (ransomware detection), OwlyScan (darknet search)
• • Lead R&D on AI for cybersecurity products; own model design, training, and integration with production systems.
• • Shipped LLM-based teaser generation pipeline (OpenAI API, LangChain) for personalized marketing content, deployed in production.
• • Built scalable web crawling infrastructure (Python, Playwright, BeautifulSoup) for darknet data extraction.
• • Designed channel-aware embedding strategies for transformer-based multivariate time series forecasting —
83% parameter reduction with no loss in accuracy (IEEE ICTAI 2025). • • Developed cross-modal anomaly detection framework (CRAFT) for cybersecurity telemetry; accepted SEC 2026. • • Built generative pipelines (GPT-2, RQ-VAE + autoregressive transformers) for synthetic tabular and time series data, improving downstream malware detector training.
Python Pytorch Machine learning Deep Learning Cybersecurity
SitInCloud & ESILV
Research Intern
avril 2023 - octobre 2023 (6 mois)
• • Implemented deep learning models (autoencoders) for ransomware detection from system I/O telemetry; integrated into OwlyShield product pipeline.
• • Built malware classification models on system behavior features; co-authored two papers (CRISIS 2024).
IRIF (CNRS & Université Paris Cité)
Research Intern
avril 2022 - août 2022 (4 mois)
• • Developed pruning optimizer for deep networks (ResNet, WideResNet) on CIFAR / ImageNet, combined with adversarial training. Supervisor: Dr. Adrian Vladu.