Irfan Muhammad Ghani2025-05-30T07:29:12+07:00
Scalable LLM Chatbot Backend
Developed a Retrieval-Augmented Generation (RAG) chatbot backend using LangChain, LangGraph, FastAPI.
Integrated tool calling, database access, and agentic LLMs.
Supported both local (Ollama, vLLM) and API-based (OpenAI, Gemini) models.
Deployed on GCP with response times under 5 seconds.
Irfan Muhammad Ghani2025-05-30T07:28:43+07:00
Document Parser & Retriever System
Built an API for parsing and retrieving documents with thousands of pages using LangChain, pdfplumber, FAISS.
Integrated agentic retriever and reranking.
Achieved ~95% retrieval accuracy in a financial regulation use case.
Irfan Muhammad Ghani2025-05-30T07:28:20+07:00
Real-Time Fraud Detection System
Implemented a semi real-time fraud detection engine with incremental learning using Scikit-learn.
Scaled dynamically using Ray for high availability and performance.