2025 · Hackathon Win
1st Place - UCLA SAIRS Hackathon
Won the sustainability track with EcoMate-AI, a multimodal agentic platform built in 24 hours.
Read the case study →
AI Enabled Data Scientist
I ship production machine learning, RAG, and LLM systems that move real business metrics - 72% lower cost on a 5M+ record/month pipeline, 27% better fraud detection, and 115x GPU speedups. Currently an MS Data Science student at USC.
Off the clock, I play table tennis and football (soccer), and love playing the piano. Always happy to talk horror movies, anime, and sports.
Work
Impact-first builds across agentic AI, GPU computing, and applied ML. Each opens a full case study with the story, demo, and links.
May 2026 - Jun 2026
Agentic AI assistant for creator teams - catches trends, decides what to film, and assigns the work. Multi-agent system with GPT-4o VLM trend detection, shipped full-SDLC through CI/CD.
May 2025 - Jun 2025
Multimodal agentic platform (Gemini 2.5 Flash) that turns receipts, images, and activity into personalized sustainability actions - built in 24 hours.
Jan 2026 - Feb 2026
Open-source CUDA library for CNN preprocessing - shared-memory tiled kernels and Python bindings delivering 115x CPU and 14x PyTorch speedups.
2025
Turns Atlassian workflow chaos into clarity with real-time flow graphs and heatmaps that reveal blockers, risks, and engagement.
Experience
Short and impact-first - full detail lives in my resume.
Liquidity AI Capital · Los Angeles, CA
Deployed production SLM inference in Docker, fine-tuning open-source models end-to-end (tokenization + LoRA adapters) for optimized, domain-specific inference on financial data.
KCC Capital Partners · Los Angeles, CA
Automated M&A target sourcing with an agentic LangChain pipeline chaining LLM entity classification, web scraping, and structured extraction to identify and rank strategic buyers.
Bajaj Finserv Health · Pune, India
Architected a production RAG pipeline (LangChain + GPT-3.5 Turbo) processing 5M+ medical records/month - cutting operational costs 72% at 92% extraction accuracy. Added Langfuse observability and Kafka event-driven inference for zero downtime.
Bajaj Finserv Health · Pune, India
Redesigned the production NER pipeline to boost fraud identification accuracy 27% with fewer false positives, and shipped a supervised model that improved operational efficiency 22%.
Bajaj Finserv Health · Pune, India
Scaled ETL and analytics to 10M+ records across 200+ features with a distributed C++/Trino/Docker system, optimized SQL, and disaster recovery.
Leadership
Building data science communities and mentoring the next generation of graduate students.
GRIDS - Graduate Data Science Club, USC
Sep 2025 - Present
Lead analytics workshops, ideathons, and projects for a 250-member graduate data science community.
USC Viterbi School of Engineering
Jan 2025 - Present
Mentor graduate students on data science career paths, technical communication, and professional development.
Education
MS, Applied Data Science · Los Angeles, CA
Jan 2025 - Dec 2026 · GPA 3.7 / 4.0
Coursework: Deep Learning, Machine Learning, Data Mining (MapReduce), Data Management.
BTech, Computer Science · Chennai, India
May 2018 - Jun 2022 · GPA 3.46 / 4.0
Coursework: Algorithms, Artificial Intelligence, Computer Vision, Probability & Queueing Theory.
Skills
Contact
The fastest ways to reach me: