Explore my Interactive 3D Portfolio with a built-in AI Assistant:
I'm a Software Developer passionate about AI Engineering and Full-Stack Development.
Outside of coding, I’m always learning, exploring new tools, looking for ways to improve both my technical skills and the overall quality of my work — or playing basketball when I need a break. 🤓
Building AI applications with:
- Large Language Models (LLMs)
- Retrieval-Augmented Generation (RAG)
- Fine-Tuning
- OpenAI API
- LangChain
- Ollama
- Python
Also experienced in:
- Next.js
- React
- Node.js
- MongoDB
A Retrieval-Augmented Generation (RAG) application that enables users to chat with PDF documents using a local Large Language Model.
Key Features
- Upload and index one or multiple PDF documents.
- Semantic document retrieval using ChromaDB.
- Context-aware question answering powered by a local LLM.
- Interactive Gradio web interface.
Built With
- Python • LangChain • Ollama • Llama 3.2 • ChromaDB • Hugging Face Embeddings • Gradio
An end-to-end LLM fine-tuning project that adapts Meta Llama 3.2 3B Instruct for customer support conversations using QLoRA and PEFT. The project covers dataset preparation, supervised fine-tuning, experiment tracking, evaluation, and LoRA adapter export for efficient inference.
Key Features
- Fine-tuned Meta Llama 3.2 3B Instruct using QLoRA.
- Parameter-efficient training with PEFT and LoRA adapters.
- 4-bit quantization using BitsAndBytes.
- Experiment tracking and evaluation with Weights & Biases.
Built With
- Python • Transformers • TRL • PEFT • QLoRA • BitsAndBytes • Hugging Face • Weights & Biases
A full-stack web application for vinyl collectors to organize, discover, and showcase their collections online.
Key Features
- Google Authentication
- Public collector profiles
- Vinyl search powered by Discogs
- Achievement and badge system
Built With
- Next.js • React • TypeScript • Tailwind CSS • MongoDB • Auth.js • Discogs API
