Skip to content
View MahdiarHarandi's full-sized avatar
😉
Focusing
😉
Focusing

Block or report MahdiarHarandi

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.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

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

Report abuse
MahdiarHarandi/README.md

Hi, I'm Mahdiar Harandi 👋

Final-year Computer Engineering student at the University of Tehran, focused on Machine Learning & NLP - building and evaluating LLM systems end to end: retrieval-augmented generation, tool-using agents, and parameter-efficient fine-tuning.

  • 🔭 Interests: LLMs, RAG & information retrieval, agentic systems, systems for AI
  • 🌱 Currently: deepening practical LLM engineering (RAG, fine-tuning, evaluation)
  • 📫 Reach me: harandimahdiar@gmail.com · LinkedIn

🚀 Featured Projects

Project What it does Highlights
Persian Legal QA with RAG Agentic RAG over scanned Persian law OCR → LanceDB (legal metadata) → LangGraph agent (rewrite · retrieve · LLM-rerank+retry · ground); RAGAS answer-relevancy ≈ 0.78
Tool-Using Travel Assistant LLM agent that plans trips with 7 tools Structured tool calling + fallbacks; 14-test suite at 100% tool routing
Instruction Tuning via LoRA PEFT fine-tuning of TinyLlama-1.1B Trains 0.2% of params; IFEval instruction-accuracy 0.09 → 0.26 (≈2.8×)

🛠️ Tech

Languages: Python · C/C++ · Java ML / NLP: PyTorch · scikit-learn · Hugging Face Transformers · PEFT/LoRA LLM / RAG: LangChain · LangGraph · LanceDB · RAGAS Engineering: Spring Boot · REST APIs · GitHub Actions · Docker · Linux


📄 Resume available on request · 🎓 Rank 305 / ~160,000 in the national university entrance exam (top 0.2%)

Pinned Loading

  1. instruction-tuning-lora instruction-tuning-lora Public

    Parameter-efficient instruction tuning of TinyLlama-1.1B with LoRA, evaluated on IFEval.

    Jupyter Notebook

  2. persian-legal-qa-rag persian-legal-qa-rag Public

    Agentic RAG over scanned Persian legal documents — OCR + LanceDB + LangGraph agent.

    Jupyter Notebook

  3. spoken-language-identification spoken-language-identification Public

    Spoken language identification from librosa audio features (KNN/SVM/MLP + clustering).

    Jupyter Notebook

  4. wifi-indoor-positioning wifi-indoor-positioning Public

    Hybrid Wi-Fi indoor positioning: k-NN fingerprinting + RSSI trilateration on ESP8266.

    C++