Rasul Alakbarli
I am an ML/AI Engineer and graduate student at Université Paris-Saclay pursuing a Master’s degree in Artificial Intelligence. My research interests span machine learning, NLP, computer vision, and pre and post training of LLMs.
Previously,I worked at Azercosmos (National Space Agency of Azerbaijan) and interned at Huawei China, where I worked on reinforcement learning training of multi-step agents.
My current focus is on distributed training of LLMs and advancements in LLM architecture.
If you are interested in collaboration or have any questions, feel free to reach out to me via email.
Education
Université Paris-Saclay
Master of Science in Computer Science
Sep 2024 - Jun 2026 | Paris, France
Specialization in Artificial Intelligence
UFAZ
Bachelor of Science in Petroleum Engineering
Sep 2020 - Jun 2024 | Baku, Azerbaijan
Focus on Oil and Gas Exploration
Experience
ML Engineer
Huawei Dongguan R&D Center
Designed and implemented a Smolagents-compatible RL training framework within rLLM which reduced integration friction for future agents. Trained and validated multi step language agents.
Computer Vision Engineer
Azercosmos
Developed aerial image segmentation/detection pipelines (U-Net/DeepLab style) and data curation workflows for satellite imagery.
Machine Learning Intern
AZAI Tech
Implemented on-device face detection/recognition using ML Kit and MobileNetV2; optimized preprocessing and model packaging for Android.
Projects
LLM Encyclopedia
- Implemented core LLM building blocks from scratch: scaled dot-product attention, multi-head attention, grouped-query attention, byte-level BPE tokenizer.
- Built models like: Transformer from ”Attention is all you need”, and a GPT-2 style decoder-only model.
Personalized Spatial Audio (HRTF Prediction) from Pinna images
- Developed an LSTM-based encoder–decoder model that infers individual HRTFs from ear images and anthropometric cues; implemented training/eval loops and metrics.