Revati Natu
AI Researcher · Builder · Forever Learner
AI Researcher · Forever Learner · Builder · Occasional Artist
I build things that learn. Researcher, builder, and sometime artist — currently deep in AI, NLP, and whatever rabbit hole opened up last night. When I'm not training models or reading papers, I'm making art with whatever's lying around, getting lost in a good book, or playing whatever I feel like on guitar — hard or not.
About
IT + AI Honours student at K J Somaiya School of Engineering, with a habit of going deep on everything I touch.
Research, code, art, music — I don't separate them. Depth over breadth, always. I write up everything I learn, build things to understand them better, and believe the best work happens at intersections.
Currently interning at Jio Creative Labs, working on AI content creation and internal tool development. Final year, finishing degree requirements in the next few months.
Projects
Interactive data analytics platform built and deployed on Streamlit — enabling real-time data exploration, visualisation, and insight generation through a clean, accessible interface.
Streamlit tool for uploading and querying research papers via Gemini Pro and LangChain — structured metadata extraction, CSV summarization, smart Q&A from PDFs.
Intelligent document analysis using RAG architecture — integrating LangChain, Ollama LLM, and vector embeddings for automated resume parsing and semantic search.
Comparative analysis of ML and DL models for cardiac arrhythmia detection on the MIT-BIH dataset, evaluating tradeoffs in precision, recall, and generalisation across model families.
Computer vision system for image forgery detection using inpainting techniques on a large public dataset — deployed with a Streamlit interface for real-time inference with confidence scoring.
Linear Stacking Classifier built by evaluating 10+ ML algorithms with advanced preprocessing. Award-winning work, presented at an international conference.
Explore the full repository — including neural-engines (Karpathy series notebooks) and open source contributions.
5 research projects carried out across medical AI, NLP & computer vision
Medical Imaging · IEEE
Framework for multi-cancer detection across five cancer types using 65,000+ histopathology images. Evaluated DenseNet201 and Vision Transformer architectures with comparative analysis across multiple metrics and validation strategies.
NLP · Mental Health AI · IEEE
Hybrid LSTM–RNN architecture trained on the Sentiment140 dataset for depression signal classification from social media text. Explores sequential modelling for mental health NLP tasks with ablation analysis.
🏆 Best Paper Award · Healthcare AI
Linear Stacking Classifier built by evaluating 10+ ML algorithms with advanced preprocessing. Presented at an international conference — earned Best Paper award for innovative contributions.
Computer Vision · Conference
Computer vision system for forgery detection using inpainting techniques on a large public dataset. Deployed with a Streamlit interface for real-time inference with confidence scoring. Presented at an International Conference.
Healthcare AI · Classification
Comparative analysis of ML and deep learning models for cardiac arrhythmia detection on the MIT-BIH dataset, evaluating tradeoffs in precision, recall, and generalisation across model families.
Experience
Jan 2026 – Present
Jio Creative Labs · Reliance Jio · Mumbai
Jan 2025 – May 2025
Commonwealth Secretariat · Remote / International
Feb 2025 – Apr 2025
Forgery Detection System via Inpainting · Independent Research
Skills
A stack built around deep learning, NLP, and generative AI — with enough fullstack to ship things end-to-end. Hover to explore.
Languages
AI / ML Frameworks
Data & Tooling
Generative AI & Creative Tools
Certifications
Achievements & Leadership
Contact
Open to research collaborations, internship opportunities, full-time roles post-graduation, and interesting conversations. If you're building something meaningful, reach out.