Skip to main content
Ventures & Projects

Building at the Intersection of AI and Finance

From AI-powered financial due diligence to campus networking — each venture solves a real problem with production-grade engineering.

Recognition

Awards and competition results

AuraHealth

🏆 Google Award

CheeseHacks 2026

TAM

🥈 2nd Place

MadData 2026

EarningsLens

⚡ Finalist

CursorHacks 2026

Active Ventures

Currently building — started 2025, ongoing

TAM

Active🥈 2nd Place — MadData 2026

Transaction Analysis Machine — AI Financial Due Diligence Platform

Demo Video

TAM — Full Product Demo

What TAM Does

TAM automates the grueling, highly manual process of Financial Due Diligence (FDD), transforming weeks of "grunt work" into seconds of automated insight. Built for Private Equity analysts and M&A professionals.

  • Quality of Earnings (QoE) — Reported vs. Adjusted EBITDA analysis
  • Net Working Capital (NWC) — Working Capital Peg calculations
  • Cash Conversion Cycle (CCC) — Operational efficiency metrics
  • Discounted Cash Flow (DCF) — 5-year projections & NPV
  • Automated PDF/Excel report generation via Playwright

AI & ML Engine

  • Document Classification (NLP) — Hybrid Hugging Face Transformers + keyword-heuristics to identify financial document types
  • Anomaly Detection (Unsupervised ML) — Scikit-Learn Isolation Forest for multivariate outlier detection across financial metrics
  • Agentic Extraction — Self-correcting extraction loop using Claude 4.5 Sonnet with statistical validation + re-examination
  • RAG Q&A — ChromaDB vector database for context-aware financial document querying

System Architecture

TAM System Architecture

Agentic Extraction Workflow

TAM Agentic Workflow

Tech Stack

frontend

Next.js 14TypeScriptTailwind CSSRecharts

backend

FastAPISQLAlchemyPydanticPython

ai

Claude 4.5 SonnetHugging Face TransformersIsolation Forest (sklearn)ChromaDB

reports

Playwright (Headless Chromium)PDF/Excel generation
Team: Yash Arvind, Diya Kothari, Hriday Thakkar·Advisors: Ben Rugg & Varun Kumar·Interest from: American Family Insurance, EY

Flux

Live on App Store4.8 ⭐ · 6 Ratings

Student Startup Co-Founder Matching · iOS Native

App Store

Designed and developed the MVP using Swift, implementing custom UI for swipe-based matching and networking. Leading a cross-functional team of six (design, development, marketing) to launch across campus. Authored business plan including market sizing, financial projections, and GTM strategy for seed-stage investments. Latest version (v2.1.0) features Google Drive resume uploads and brand new landing page.

SwiftiOS NativeFirebaseProduct ManagementGTM Strategy

6

Team Members

v2.1.0

Latest Version

4.8 ⭐

App Store Rating

App Store Reviews

"Cofounder"

★★★★★

"Found my co-founder on this app in my uni… we were the perfect connection and we are now in the product development stage!"

Vashishth S · Nov 2025

"Love the App"

★★★★★

"Simple, easy, and straightforward way to connect with talent entrepreneurs!"

CTC-WI · Oct 2025

"Useful"

★★★★★

"Helped me find a co founder for my comp sci project"

MysteryCoder456 · Oct 2025

Hackathon Projects

Shipped under pressure — all built in 2026

Hackathon
🏆 Google Award — CheeseHacks 2026

AuraHealth

AI Patient Follow-Up Agent

Won the Google Award at CheeseHacks 2026. AI-powered healthcare assistant that calls patients post-consultation via Twilio WebSocket streaming, Google STT/TTS, and Gemini 2.5 Flash. Uses Pinecone RAG for consultation context and Cloud Scheduler for automated follow-ups. High-urgency triage triggers SMS alerts to doctors.

CheeseHacks 2026

Event

FastAPINext.jsWebSocketsGeminiTwilioPineconeGoogle Cloud
Hackathon
⚡ CursorHacks 2026

EarningsLens

Real-Time Earnings Call Deception Detection

Built at CursorHacks 2026. Analyzes earnings call transcripts using Claude AI with forced tool use and an 11-signal deception taxonomy (hedging, deflection, overclaiming, etc.). Cross-references against SEC filings via ChromaDB vector search. Tracks per-speaker credibility scores and pulls live market data via Finnhub.

CursorHacks 2026

Event

ReactFastAPIClaude AIChromaDBSEC EDGARFinnhub
Hackathon
⚡ ClaudeHacks 2026

HelloNeighbour

Voice-Based Anonymous Dorm Connection App

Built at ClaudeHacks 2026. Voice-based anonymous neighbor connection app scoped for UW-Madison dorms. Features KNN (k=3) matching algorithm from scratch, Claude API for profile analysis and meetup suggestions, Web Audio waveform visualization, and 15-second voice notes. Deployed to Vercel.

ClaudeHacks 2026

Event

Next.jsClaude APIKNN (from scratch)Web Audio APIVercel

Past Projects

Completed work and earlier ventures

Completed

BoxMate

Student Storage Marketplace

Led branding, platform design, and customer outreach strategies. Generated $10,000 in revenue within the first 2 weeks. Performed regression analysis to optimize pricing strategy and built forecasting model using Python.

$10K in 2 weeks

Revenue

PythonPricing OptimizationForecastingOperations
In Progress

AI Hedge Fund

Autonomous Trading System

Building automated AI-agent-driven infrastructure with strategy execution, portfolio rebalancing, and risk controls. $100K+ portfolio using n8n, Python trading engine, and ML models.

$100K+

Portfolio

n8nPythonMLQuantitative Finance
Completed

Stock Trading System

Algorithmic Trading Engine + LLM

Full-stack platform integrating PAMR (Passive Aggressive Mean Reversion) algorithm with LLM-based market commentary. Next.js frontend, PostgreSQL, Python engine, Marketstack API.

Next.jsPythonPostgreSQLPAMRLLM APIs

Financial Expertise

Core competencies developed through experience and execution

Due Diligence

  • Financial statement analysis
  • Red-flag identification
  • Balance sheet deep-dives
  • Background checks & IAR reports

Valuation & Modeling

  • DCF modeling
  • Comparable company analysis
  • Financial projections
  • LBO analysis basics

Data & Analytics

  • Predictive modeling
  • Revenue forecasting
  • Demand analysis
  • Regression optimization

Automation

  • Automated PDF/Excel generation
  • n8n workflow automation
  • Dashboard creation
  • NLP document parsing

Technical Stack

Languages & Frameworks

Languages

PythonSQLRSwiftTypeScript

Frameworks

FastAPIReactNext.jsLangChainPandasNumPyScikit-learn

Tools & Specializations

Tools

TableauExceln8nGitChromaDBPinecone

Specializations

Agentic AINLPQuantitative FinanceML/Predictive Analytics

Interested in Collaborating?

Always open to discussing new ventures, technical challenges, or investment opportunities.

Get in Touch