Saran Bluri

SARANBLURI

Explore
About

The story so far.

Saran Bluri

Hey there! I'm Saran Bluri (Bodduluri). I'm currently pursuing an MS in Mathematics in Finance at the NYU Courant Institute, sitting at the intersection of quantitative finance, artificial intelligence, and the startup ecosystem.

Since moving to the US, I've thrown myself into the builder community, hitting about 8 hackathons in just the past couple of months. I absolutely love building great products, and talking to people. Right now, I'm channeling that energy as a founder, actively building out my own startup.

My technical foundation stems from a background in CS and Machine Learning from Shiv Nadar University, along with 1.5 years developing statistical arbitrage and systematic strategies as a Quant Analyst at PL Capital and True Beacon.

Whether I'm teaching AI concepts, creating content across YouTube and Substack, or hitting the gym, my core focus is pushing the boundaries of what technology can achieve. We are standing on the brink of a massive shift in how humanity interacts with information and machines, and the potential for positive impact is boundless. I am incredibly honored and excited to be building at this time, and I am fully committed to shaping that future.

Research

Recent
Publications.

Kovvuri Sai Gopal Reddy, Saran Bodduluri, A. Mudit Adityaja, Saurabh Shigwan, Nitin Kumar, Snehasis Mukherjee (2024). UnSeGArmaNet: Unsupervised Image Segmentation using Graph Neural Networks with Convolutional ARMA Filters. In 35th British Machine Vision Conference 2024, BMVC 2024, Glasgow, UK.
Balakrishnan, R. M., Bodduluri, S. (2024). Primary Market Conditions and Technical Analysis. In Proceedings of the 9th PAN IIM World Management Conference.
Work

Things I've built.

InstaMarket

Multi-Agent AI Trading Platform

A multi-agent AI simulation platform that eliminates friction in trading news. Features an MV3 Chrome extension backed by AWS Bedrock Nova Lite multi-agent workflow. Ingests live Polymarket odds and multi-source dossiers via Apify to simulate market reactions, executing trades directly via Solana RPC.

AWS BedrockSolanaChrome ExtensionApify

Sese

Multimodal AI Study Companion

A multimodal AI study companion providing real-time, voice-guided coaching for handwritten problem-solving. Uses computer vision to parse math/handwritten input and generates interactive feedback. Combines RAG-based textbook retrieval with an adaptive voice agent.

Computer VisionRAGVoice AIOneNote

clauDuck

AI Academic Companion

An AI academic companion integrated directly into Canvas and Ed. Powered by Claude, it provides pre-class lecture priming, interactive slide queries during lectures, post-class quizzes, and uses Linkup to surface relevant campus and NYC events based on coursework.

ClaudeCanvasRAGLinkup

Claworporate

AI Trading Desk for Prediction Markets

An AI trading desk for prediction markets that lives inside group chats. A multi-agent system analyzes real-time news, social signals, and market data to estimate edge and coordinate trades. Built on XMTP infrastructure with Polymarket data.

Multi-AgentXMTPPolymarket

DROBE

Agentic Fashion Discovery

A personalized agentic shopping tool that turns fashion discovery into dialogue. Built using a RAG-based multimodal retrieval system, powered by CLIP embeddings for product images and a FAISS vector database.

CLIPFAISSRAGMultimodal

FoodPo

Voice-First Food AI Agent

A voice-first AI agent for food decisions. Users speak constraints ("veg, high protein, under $15") and it ranks options across DoorDash, Uber Eats, and Grubhub with true price and macro calculations.

Voice AIMulti-platformLLM
UnSeGArmaNet

UnSeGArmaNet

Image Segmentation Model

Unsupervised image segmentation model using Graph Neural Networks with ARMA filters which beats SoTA.

PyTorchGNNComputer Vision
SokoRamp

SokoRamp

AI Sokoban Solver

A Python-based AI Sokoban Solver utilizing the greedy A* algorithm with a Goal Pull Distance heuristic.

PythonNumpyA* Algorithm
CARMA

CARMA

Crime-Surveillance

AI-powered video surveillance system that detects violent activities in real-time using advanced LRCN and ConvLSTM models.

PythonLRCNConvLSTM
The Efficient Frontier

The Efficient Frontier

Markowitz Portfolio Theory

Exploration of the efficient frontier of a multi-stock portfolio using Modern Portfolio Theory.

PythonTableauVisualization
S&P Hangman

S&P Hangman

Hangman Game

S&P 500 themed Hangman game built using Python.

Python
Gallery

Showcase.

BMVC 2024

BMVC 2024

Presenting · Glasgow

Paper Presentation

Paper Presentation

Research · Conference

On Stage

On Stage

Speaking

Claude Hackathon

Claude Hackathon

Columbia × NYU

YC Event

YC Event

London, UK

Y Combinator

Y Combinator

SF · 2026

Graduation

Graduation

SNU · 2024

ClawHacks

ClawHacks

Finalist

Grand Canyon

Grand Canyon

Arizona

Oakland Bridge

Oakland Bridge

New Year's · SF

Miami

Miami

Florida

Lotus Temple

Lotus Temple

New Delhi

Mumbai

Mumbai

Skyline · India

Family

Family

Home

Friends

Friends

Together

Beach Day

Beach Day

With Friends

Graduation

Graduation

Ceremony

Contact

Get in touch.

It's an incredible time to be building the future. Whether you want to talk about AI, the startup ecosystem, or just exchange creative ideas, my inbox is always open.