💡 Click any section header below — most of them expand into more detail.
role: Researcher
position: Research Assistant, EliteLab AI
focus: Intelligent AI systems for domain-specific applications
mission: >
Building AI systems that can understand complex legal language,
empower low-resource communities, and make knowledge more accessible.
|
|
📖 IndicIPR-QA — Extractive QA over Indian Intellectual Property Law 
A benchmark dataset + extraction pipeline for answering questions directly from Indian IP law text — patents, trademarks, and copyright statutes — built to make dense legal language machine-queryable.
⚖️ Legal Question Answering — Domain-specific QA over legal corpora 
End-to-end QA systems tuned specifically for legal text, where precision and traceability to source clauses matter as much as the answer itself.
🕸️ Legal Knowledge Graph — Structuring statutes, cases, and precedent
Converting unstructured legal documents into queryable knowledge graphs that connect statutes, case law, and precedent for downstream reasoning.
🔍 Synthetic DAPT — Synthetic Domain Adaptive Pretraining for Extractive QA 
Using synthetically generated domain data to adapt pretrained language models for extractive QA — cutting the cost of scarce, expensive expert-annotated data.
🎙️ BanglaMix — Benchmark for Code-Switching ASR in Bengali Dialects
A speech recognition benchmark built for the messy, real reality of Bengali speech — dialect variation and code-switching included — where most ASR systems quietly fail.
Hugging Face Transformers Sentence Transformers LangChain FAISS Ollama Whisper spaCy NLTK pandas NumPy Matplotlib
🏆 GitHub Trophies
- Publish IndicIPR-QA benchmark
- Build first legal knowledge graph prototype
- Release BanglaMix dataset publicly
- Finish Synthetic DAPT paper revisions
- Open-source the legal QA pipeline
- Collaborate on a cross-lingual legal AI project 👀
🤖 Large Language Models & RAG
Fine-tuning, retrieval-augmented generation pipelines, and adapting LLMs to specialized, high-stakes domains.
⚖️ Legal AI & Legal Text Analytics
Applying NLP to statutes, case law, and IP documents — extraction, structuring, and reasoning over legal language.
🕸️ Knowledge Graphs & Information Retrieval
Turning unstructured text into structured, queryable graphs, and building retrieval systems that scale to real document collections.
🌍 Low-Resource Languages & Bengali ASR
Speech and text technologies for dialects and languages that mainstream models tend to ignore — code-switching included.
Working on legal AI, RAG, or low-resource NLP? Let's talk.
📧 Email: niloy8649@gmail.com 🌐 Website: https://niloycste.github.io/niloycse/ 🤗 Hugging Face: https://huggingface.co/niloycste68
"Building AI systems that can understand complex legal language, empower low-resource communities, and make knowledge more accessible."