Skip to content
View behroozazarkhalili's full-sized avatar
πŸ‘¨β€πŸ’»
I'm developing the DL & ML Code.
πŸ‘¨β€πŸ’»
I'm developing the DL & ML Code.

Highlights

  • Pro

Block or report behroozazarkhalili

Block user

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.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
behroozazarkhalili/README.md

Hi there, I'm Behrooz Azarkhalili πŸ‘‹

πŸ”¬ AI Scientist & Engineer β€’ ML/DL & LLM Fine-Tuning Polyglot β€’ Multimodal Maverick

With nine years of deep immersion, I architect, train, and optimize state-of-the-art ML, DL, and LLM systems β€” spanning NLP, computer vision, speech, and multimodal architectures that truly understand the world.

βš™οΈ Core Expertise

🧠 LLM Fine-Tuning

  • Parameter-Efficient Techniques: Implement LoRA, QLoRA, Prefix-Tuning, and Adapter-Tuning using Hugging Face PEFT library to inject lightweight adapters into frozen transformers, cutting memory & parameters by orders of magnitude while maintaining full-scale performance
  • Quantization & Optimization: Leverage bitsandbytes + 8-bit/4-bit quantization for efficient training and QLoRA benefits on consumer hardware
  • Advanced Training: Use TRL and Unsloth for instruction-tuning and RLHF workflows, including DPO, GRPO, PPO, and others

πŸ”¬ Multimodal / Classic ML & DL

  • Neural Architectures: Build custom transformers, CNNs, and RNNs across text, vision, and audio modalities
  • Research & Development: Contribute to research & deployment projects at UC Berkeley, Georgia Tech, LAVIVO, and AI VIVO

πŸ”„ Full ML Lifecycle

  • End-to-End Pipelines: Data ingestion β†’ experimentation (MLflow, W&B) β†’ scalable inference (AWS SageMaker, Gradio, Streamlit)

πŸ› οΈ Technical Stack

πŸ’» Programming Languages Python R Scala MATLAB Octave
πŸ€– ML/DL Frameworks TensorFlow PyTorch Keras Scikit Learn
πŸ”€ NLP Transformers spaCy
🧠 LLMs API Claude OpenAI Deepseek Gemini LangChain Ollama
βš™οΈ LLMs Fine-tuning TRL PEFT Unsloth
⚑ MLOps MLflow W&B Apache Spark Apache Airflow
πŸš€ Deployment & Cloud AWS Gradio Streamlit Heroku
πŸ—„οΈ Databases MongoDB SQLite PostgreSQL Weaviate Qdrant
πŸ› οΈ Development Tools Git GitHub GitLab
πŸ’» IDEs VS Code PyCharm Colab RStudio
πŸ“ˆ Productivity LaTeX Markdown WakaTime ClickUp
πŸ’» Operating Systems Ubuntu Linux Mint Debian

πŸ“ͺ Contact Me

Twitter logo Gmail logo

πŸ“š Publications

Google Scholar logo Arxiv logo

I'm a Night πŸ¦‰

🌞 Morning    49 commits     β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   25.13% 
πŸŒ† Daytime    36 commits     β–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   18.46% 
πŸŒƒ Evening    76 commits     β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   38.97% 
πŸŒ™ Night      34 commits     β–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   17.44%

πŸ“… I'm Most Productive on Tuesday

Monday       8 commits      β–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   4.1% 
Tuesday      75 commits     β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   38.46% 
Wednesday    41 commits     β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   21.03% 
Thursday     13 commits     β–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   6.67% 
Friday       9 commits      β–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   4.62% 
Saturday     9 commits      β–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   4.62% 
Sunday       40 commits     β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   20.51%

πŸ“Š This Week I Spent My Time On

πŸ’¬ Programming Languages: 
Python                   21 hrs 21 mins      β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘   90.91% 
JSON                     39 mins             β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   2.81% 
Markdown                 33 mins             β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   2.36% 
Git Config               22 mins             β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   1.59% 
Other                    16 mins             β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   1.16%

πŸ”₯ Editors: 
VS Code                  23 hrs 30 mins      β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ   100.0%

πŸ±β€πŸ’» Projects: 
dspy                     6 hrs 38 mins       β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   28.29% 
ALTI-XAI                 4 hrs 44 mins       β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   20.19% 
cookbook                 3 hrs 40 mins       β–ˆβ–ˆβ–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   15.61% 
Milvus-RAG               2 hrs 28 mins       β–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   10.53% 
attachments              2 hrs 4 mins        β–ˆβ–ˆβ–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘β–‘   8.8%

πŸ’» Operating System: 
Linux                    23 hrs 30 mins      β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ   100.0%

Last Updated on 02/09/2025

Pinned Loading

  1. unslothai/unsloth unslothai/unsloth Public

    Fine-tuning & Reinforcement Learning for LLMs. πŸ¦₯ Train OpenAI gpt-oss, Qwen3, Llama 4, DeepSeek-R1, Gemma 3, TTS 2x faster with 70% less VRAM.

    Python 45.7k 3.7k

  2. huggingface/trl huggingface/trl Public

    Train transformer language models with reinforcement learning.

    Python 15.6k 2.2k