Soumik Deb Niloy

Soumik Deb Niloy

AI / ML Engineer · Research Assistant · NLP & Computational Biology

d19niloy@gmail.com · 🔗 LinkedIn · 💻 GitHub · 📞 +880 17963 11815
📄 Download CV

About

Research-driven AI/ML engineer focusing on deep learning and graph methods for biological data, medical imaging, and multilingual NLP. Experienced in PyTorch, TensorFlow, graph neural networks, and building end-to-end research prototypes.

Current: Research Assistant — BRAC University & Center for Computational and Data Sciences (2025–present)

Technical Skills

PythonPyTorchTensorFlow PyTorch GeometricHuggingfaceOpenCV Scikit-learnLangChain FAISS / MilvusDocker / Git / LaTeX

Experience

Research Assistant — BRAC University

Conducting AI/ML research on peptide classification, gene expression, and medical imaging using GNNs and deep learning frameworks.

Research Assistant — IUB (CCDS)

Developing computational models for biomedical and NLP data; conducting experiments and literature review for journal publication.

Projects

Customized EfficientLiteNetB0Hybrid CNN+Transformer

Lightweight model for video frame analysis (0.7 GFLOPs, 5 MB). Includes preprocessing, Grad-CAM visualization, and edge-device inference pipeline.

PyTorchOpenCVGrad-CAM
AI-Driven Financial Fraud DetectionGNN · Neo4j

Transaction-graph GNN for fraud detection, real-time anomaly alerts, and explainability visualization in Neo4j.

PyGDGLXGBoost
Second Brain — Personal Knowledge CuratorRAG · Streamlit

Local pipeline that ingests PDFs/URLs, creates embeddings with SentenceTransformers, stores them in Chroma, and enables RAG queries via FastAPI.

FastAPIChromaSentenceTransformers
Python Code Generation from BanglaLLM · Agentic System

Bangla instruction-to-Python generator/checker/debugger pipeline using local LLMs and prompt optimization.

OllamaPrompt Engineering
OCR — Handwritten PrescriptionPaddleOCR

Multilingual OCR system with post-processing for drug-name verification and web/mobile upload interface.

PaddleOCRNLP

Honors & Achievements