Tanya Nair

Johns Hopkins Electrical and Computer Engineering

Undergraduate at Johns Hopkins, majoring in Computer Engineering. Passion for applying cutting-edge AI and emerging technologies to problems spanning medicine, accessibility, and more. Interested in making a sustainable, accessible, AI-enhanced future.

Core Skills

Languages: C, C++, Python, Java
Frameworks: PyTorch, TensorFlow, HF Transformers
Systems: Linux/UNIX, Git, Docker, Cloudflare Workers, Render
Focus Areas: Sustainable Systems, Accessibility Tech, Healthcare & Surgical AI, Resource Optimization

My Areas of Interest:

RAG and Knowledge-Grounded LLMs
Energy- and Resource-Efficient AI Systems
Multilingual and Cross-Cultural Language Models/Information Flows
Societal, Ethical, and Accessibility Impacts of AI
LLM Reliability & Security: Hallucinations, Robustness, and Data Privacy

Current Experience

Johns Hopkins Center for Language and Speech Processing: Working with Prof. Ziang Xiao and Nikhil Sharma to build a benchmark for multilingual LLM outputs to queries. Continuation of work by Sharma, et. al..
ARGOS Research Lab: Transparency Gaps in Surgical LLM Research, Geospatial Disparities in Community Opportunity Metrics: A Data-Driven Framework for Equitable Surgical Resource Allocation
University of WA: Working on an automated pipeline to enhance accessibility of .pptx presentations for vision-impaired students. Presented at EDUCAUSE 2025. Contribute to the open source script!

Past Experience

Dartmouth Health: Built a RAG model for clinician cross-referencing of pathology reports with fusion vector search (Okapi BM25 and FAISS), implemented Microsoft's phi-2 for efficiency. Link to project portfolio website here.
Published Research: Natural Language Processing used in sentiment analysis of poetry: a study of six common techniques
Won Hackathons: HopHacks (1st Place Data Visualization Track), HackPNW (Best Use of Sponsor's API - Twilio)

Additional Projects

pokedex-sql: A SQL-powered Pokédex dashboard for exploring Pokémon data by generation, stats, and type distributions. Built on a normalized SQLite schema with Python-backed analytics and a Streamlit frontend. Features generation-based filtering, stat aggregation/ranking, and a Pokédex-style sprite list.
Spark Accelerator: Building a research-informed life analytics platform.
Reading: 30 books this year!
YouTube channel/Blog: Posting about AI & linguistics once a month (in an ideal world).