NASA Space Apps Challenge 2024 — Regional Winner & Global Nominee
I'm an AI Engineer at Softvence building production-ready AI systems powered by LLMs, Retrieval-Augmented Generation (RAG), and autonomous agent pipelines.
My focus is designing scalable AI architectures, including multi-agent orchestration, semantic retrieval systems, and document intelligence pipelines.
I enjoy solving problems in low-resource language AI, particularly Bangla NLP, and building systems that work reliably in real-world environments.
📍 Dhaka, Bangladesh
🌐 Portfolio: https://mrshibly.onrender.com
📫 Email: mahmudurrahman858@gmail.com
- LLM Systems & Transformer Architectures
- Multi-Agent AI Systems
- Retrieval-Augmented Generation (RAG)
- Semantic Search & Vector Databases
- Document AI & OCR Pipelines
- Low-Resource NLP (Bangla Language AI)
Multi-agent LLM system that autonomously searches academic papers, analyzes PDFs, and generates structured research reports.
Tech:
FastAPI FAISS SentenceTransformers Docker WebSockets
Semantic retrieval system over large legal and economic documents with grounded answer generation.
Tech:
FAISS Flan-T5 FastAPI Transformers
OCR + LLM pipeline that converts multilingual documents into structured JSON outputs.
Tech:
EasyOCR OpenCV Groq Llama 3 FastAPI
Training a GPT-style decoder-only Transformer from scratch including tokenizer, dataset pipeline, training loop, and text generation.
Tech:
PyTorch Transformers Tokenization Mixed Precision Training
- 🏆 NASA Space Apps Challenge 2024 — Regional Winner
- 🥇 Back-to-Back Champion — Google Hacking Contest (DIU)
- 💻 Solved 100+ competitive programming problems
- 🧠 ICPC Preliminary Contestant (2021, 2022, 2024)