Skip to content

AI StudentSoftware EngineerBuilder

AI @

Location
West Lafayette, IN
Weather
Local time
Last listened to
Status
Software engineering — co-building at MergeWorks, shipping Hunch

I'm an AI student at Purdue. Most of what I know came from building things that started as ways to make my own life easier.

Before code, it was badminton. I was ranked No. 1 in the country.

Portrait of Srijan Challapalli
Based in
West Lafayette, Indiana
Studying
B.S. Artificial Intelligence, Purdue · '28
Off-screen
Badminton, lifting, hiking, football
Open to
SWE & AI engineering internships

  1. 2026

    May 2026 — Present

    MergeWorks — Software Engineer Intern

    Built the M&A due-diligence pipeline that turns messy financial packets into source-cited, audit-ready analysis.

    p50 per packet (125s p95), down from hours
    71s
    p50 per packet (125s p95), down from hours
    passing unit tests
    111
    passing unit tests
    legacy workflows retired
    6
    legacy workflows retired

  2. 2025

    May 2025 — Aug 2025

    Cosmos Granite & Marble — Software Engineer Intern

    Built the lead-generation pipeline behind the sales team's outreach across target states.

    prospective customers found
    1,000+
    prospective customers found
    sales
    +18%
    sales
    outreach click-through
    +15%
    outreach click-through

Stack

What I reach for

Languages

Python / TypeScript / Rust / C++ / Swift / Java / SQL

Frameworks

React / Next.js / FastAPI / Flask / Express / SwiftUI

Data & tools

PostgreSQL / Supabase / Pandas / XGBoost / Linux / Vercel / Vitest

Interests

AI Systems / Distributed Systems / Full Stack / Developer Tools

Now

What I'm exploring · Oct 2026

  1. 01Distributed systemsRaft, replication, and what happens when a node lies.
  2. 02AI inferenceBatching, KV caches, and why tokens/sec isn't the whole story.
  3. 03Rust & C++Keeping two LSM engines byte-compatible.
  4. 04Blockchain infrastructureOracles, AMMs, and contract-size limits.
  5. 05Neural networksRebuilding the basics from scratch, without a framework.

(04)
ChitYap profile screen: “Make your momentum visible.”
ChitYap friends screen: “Made for lasting friendships.”
ChitYap challenges screen: “Build streaks together.”

012025iOS · Co-founder

ChitYap (opens in a new tab)

A privacy-first social accountability app for close friends.

Co-founded and shipped to 300+ beta users. Real-time feeds, offline retry, push, and an on-device Apple Vision rep counter — on a Supabase/Postgres backend locked down by 60+ row-level security policies.

300+ beta users · ~5,000 posts · 60+ RLS policies

Swift / SwiftUI / Supabase / PostgreSQL / Apple Vision / HealthKit / APNs

Website(opens in a new tab)App Store(opens in a new tab)Private repo

022026Systems · Rust & C++

LSM-Tree Storage Engine (opens in a new tab)

A persistent key-value store written from scratch — then ported to C++.

Write-ahead log, memtable, SSTables, Bloom filters, background compaction, and crash recovery via WAL replay. The C++17 port writes SSTables byte-compatible with the Rust version.

WAL replay · p99 via Criterion + perf · Rust ⇄ C++17

Rust / C++ / Tokio / Criterion / CMake / Linux perf

GitHub(opens in a new tab)

032026On-chain · Full Stack

Hunch (opens in a new tab)

A prediction market where every YES/NO share is fully collateralized by ETH.

Anyone opens a market, seeds it with ETH, and trades the outcome on a constant-product AMM — held solvent by an invariant that pins the contract's balance to shares outstanding. A crash-proof Postgres indexer keeps the books and recovers without double-counting.

7,500 simulated ops · 0 solvency failures · p99 ≤ 118ms indexer

Solidity / Hardhat / TypeScript / Next.js / PostgreSQL

GitHub(opens in a new tab)
DDillon AIby MergeWorks
Dillon AI deal memo: Apex Industrial Technologies, proceed with conditions, valuation range $10.8M–$15.2M
Deal analysis scores: overall 54, valuation 65, cash flow 68, risk 20, growth 60

042026AI Agents · Finance

Financial Due Diligence Agent

An AI system for Quality of Earnings work that scores how far each source can be trusted.

Weighs evidence on recency, corroboration, and materiality, flags contradictions, and analyzes revenue concentration — while every number is checked deterministically, never by the model.

5,000+ agent runs · 71s p50 per packet · 0 LLM arithmetic

TypeScript / LLM agents / PostgreSQL / Supabase / LlamaParse

Built at MergeWorks

Contact

srijanchallapalli@gmail.com
GitHub(opens in a new tab)LinkedIn(opens in a new tab)X(opens in a new tab)Resume(opens in a new tab)