Data Science student and ML engineer based in Dhaka, Bangladesh.
Building RAG systems, LLM agents, and memory infrastructure. Interested in mechanistic interpretability.
- Associate Product Engineer, DeepDive at Markopolo AI, an AI social listening and narrative intelligence platform
- Research Assistant at United International University, working on autonomous agents and multi-agent systems
- B.Sc. in Data Science, United International University
Previously Jr. Software Engineer at AskTuring.ai, where I owned the Eval Service and admin dashboard and architected Deepen Answer: a production RAG pipeline combining extended-thinking LLMs, an agentic gap-fill layer, and hybrid BM25 + vector search with reciprocal rank fusion.
fizzlabai · Deep research agent running a deterministic 8-phase pipeline over the open web, returning cited Markdown with a full execution trace. FastAPI, LangGraph, Postgres/pgvector, OpenTelemetry.
Engram · Persistent memory infrastructure for AI agents. Hybrid search, graph traversal, and durable task ledgers. Cuts LLM context cost up to 22x while holding 90%+ accuracy on LongMemEval, LoCoMo, and BEAM.
Personal Knowledge Assistant · Multi-agent RAG with confidence-based routing, knowledge-strip decomposition, and automatic web-search fallback.
Analyzia · LangChain agent that turns plain-English questions over CSV data into pandas analysis and Plotly visualizations.
SmartPrice Intelligence · 10,000+ smartphone listings scraped across 37 brands, then CatBoost/XGBoost models on 50+ engineered features (R² = 0.772).
FizTorch · NumPy deep learning framework built from scratch, with an autograd engine, core layers, and an Adam optimizer.
Also: LangGraph · LlamaIndex · CrewAI · vLLM · XGBoost · CatBoost · pgvector · FAISS · Pinecone · Weaviate · MLflow · Weights & Biases · DVC
Champion, IC6 University Innovation Hub Program (BDT 60,000 pre-seed) · Best Prototype, Blockchain Olympiad at UIU CSE Fest 2025 · Champion, UIU CSE Project Show Spring 2024 · 1st Runners Up, Data Innovators Challenge 2024