Block or report baslia
Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.
You must be logged in to block users.
Contact GitHub support about this user’s behavior. Learn more about reporting abuse.
Report abuseScaling LLMs & Autonomous Agentic Workflows at Production Grade
🎯 11+ years leading AI teams across 3 continents
🤖 25+ engineers directed across cross-functional teams
💰 $12MM+ revenue impact via AutoML platform
🏭 60,000+ models deployed to production globally
🔄 82% manual labeling replaced by agentic workflows
🎓 École Centrale Paris · MIT Sloan (Visiting)
🌍 US Green Card Holder & EU Citizen
I build and scale production-grade AI systems — from fine-tuning foundation models (RLHF/DPO/SFT) to orchestrating multi-agent workflows with LangGraph and MCP. My focus is bridging the gap between research and high-impact business automation.
Current obsessions: Agentic AI · LLM-as-a-Judge · Responsible AI · Synthetic Data Generation
| GenAI & LLMs | Infrastructure | Languages & Data |
|---|---|---|
| LLM Fine-tuning (RLHF/DPO/SFT) | Vertex AI · SageMaker | Python (Expert) |
| RAG · LangGraph · MCP | Kubernetes · Docker | SQL · BigQuery · Snowflake |
| LoRA/QLoRA · Quantization | MLflow · CI/CD | PyTorch · JAX · TensorFlow |
| Prompt Engineering | Pinecone · FAISS · Elasticsearch | Apache Spark · Pandas |
| Gemini · Llama · Claude · GPT-4 | vLLM · DeepSpeed · FSDP | Hugging Face Transformers |
|
🤖 Agentic Orchestration Designed multi-agent workflows automating 82% of manual labeling, improving processing speed by 35% |
🛡️ Responsible AI Architected "LLM-as-a-Judge" framework for hallucination detection — 99.9% safety compliance |
|
💡 AutoML Platform Built platform generating $12MM in incremental revenue with 60,000+ production models |
⚡ Edge Optimization Led quantization (GGUF/FP8) and distillation for on-premise deployment under strict privacy constraints |
|
🔍 Multimodal Search Integrated CLIP-based embeddings with Pinecone, reducing search latency by 60% |
🧬 Synthetic Data LoRA-finetuned generation engine boosting model coverage by 45% for low-resource categories |
Open Food Facts — Lead AI Contributor
- Architected a VLM-based OCR system for 3M+ product images (300% ingestion speed improvement)
- Built deep learning classifier for Nutri-Score prediction from ingredient lists
Community
- 🏅 Winner — French Government Food Data Hackathon (2024)
- 🧑⚖️ Judge — Devpost, AiGoLearning, NextAI, Hackathon Raptors (2025)
- 🎓 Trainer — EUROSAE: "The Renewal of Artificial Intelligence" (2024)
| Institution | Focus | |
|---|---|---|
| 🇫🇷 | École Centrale Paris — MSc Engineering | Quantitative Research, ML, Applied Math |
| 🇺🇸 | MIT Sloan — Visiting Student | AI, Statistical Learning, Generalization Theory |
Certifications: AWS ML Specialty (910/1000) · AWS Solutions Architect · CAPM (PMI)
- Generalization Bounds for Learning with Linear and Quadratic Side Knowledge — MIT · arXiv:1405.7764
- Batch Process Monitoring by DTW and K-means Clustering — AIChE Annual Meeting 2015
🌐 French (Native) · English (Fluent) · German (Intermediate)
Let's build something remarkable together.
Pinned Loading
If the problem persists, check the GitHub status page or contact support.