I bridge the gap between raw data and live intelligence, building AI systems from the first byte to the final decision.

• Design end to end AI systems (data → model → inference → monitoring)
• Ship powerful ML systems: CV pipelines, RAG systems, RLHF, and forecasting
• Deploy across edge, embedded, and cloud with latency & cost constraints
• Translate research ideas into robust, and scalable products

Real-time multi camera recognition, face quality scoring, multi frame verification, and personnel registry.
Tech: PyTorch, YOLOv8–v11, InsightFace, OpenCV, FastAPI

Domain specific retrieval + MT, dense retrieval with Sentence Transformers & ChromaDB, PPO based RLHF.
Tech: HuggingFace, LangChain, ChromaDB, PPO

Viral product detection, time to peak forecasting, multi-horizon models using cross platform signals.
Tech: TFT, Transformers, LightGBM

Product syncing, sourcing automation, logistics & fulfillment tracking, $1M+ revenue impact in one year.
Tech: Next.js, Spring Boot

Real-time fleet monitoring with protocol decoding and high throughput ingestion.
Tech: TCP servers, protocol parsing, backend systems

Computer Vision : Detection, tracking, camera calibration, face quality, real-time optimization.
LLMs & NLP : RAG (dense + hybrid), prompt engineering, RLHF, multilingual/low resource NLP.
Time-Series : Seasonality & trend modeling, multi-horizon & probabilistic forecasting.
Systems & Deployment : GPU/MPS memory tuning, latency/throughput tradeoffs, edge & cloud rollout, fault tolerance.

• Experiment tracking & model versioning (MLflow / custom)
• CI/CD for ML, containerized inference
• Monitoring & observability (logs, metrics, alerts)
• Secure data pipelines & feature validation

• Efficient multimodal fusion (vision + audio + text)
• Retrieval reranking & contrastive retrieval improvements
• Real-time speech-to-text / low-latency TTS
• Scalable forecasting across multi-country marketplaces

• Email: hello@badr-mellal.com
• Portfolio: https://badr-mellal.com

Building AI systems that actually ship.