|
Machine Learning Engineer passionate about Language Model, Computer Vision and software engineerin Currently:
|
|
Each one is a system, not a script. Built to understand, not to show.
| # | Project | Category | Description | Stack |
|---|---|---|---|---|
| 01 | 🔍 RAG Pipeline | RAG |
Offline PDF Q&A with local inference. Air-gapped, zero cloud dependency. Semantic retrieval with tuned chunking, embedding model selection, and re-ranking. | LangChain · ChromaDB · FAISS · Ollama · FastAPI · Docker |
| 02 | 📄 LLM Document Q&A | LLM |
Multi-document reasoning on Llama 3.2. Handles heterogeneous formats, long-context windows, and cross-document synthesis without naive truncation. | Llama 3.2 · HuggingFace Transformers · LangChain · FastAPI |
| 03 | 🗄️ Text-to-SQL | NLP |
NL → production-grade SQL with self-correction loop. Schema-aware prompting, semantic error detection, handles aggregations and nested queries. | SQLCoder · LangChain · PostgreSQL · FastAPI · Docker |
| 04 | 📚 Semantic Book Recommender | NLP |
Embeddings + emotion-aware filtering. Captures semantic content, emotional tone, and thematic resonance separately — combined at query time. | Sentence-BERT · FAISS · Gradio · Pandas |
| 05 | 🧠 ML Papers From Scratch | Research |
Transformers, ResNet, BERT — implemented from equations, not checkpoints. Attention, positional encoding, residual connections, masked language modeling. | PyTorch · NumPy · Math |
| 06 | ⚙️ ML/DL From Scratch | ML |
Backprop, gradient descent, SVMs, decision trees, k-means — zero scikit-learn, zero autograd. If you can implement it from scratch, you understand it. | NumPy · Pure Math |
Core ML
Infrastructure
Languages
╔══════════════════════════════════════════════════════════════╗
║ ║
║ "The goal is not to use the latest model. ║
║ The goal is to solve the problem well." ║
║ ║
╚══════════════════════════════════════════════════════════════╝
