FinTech Systems Engineer

Building low-latency trading infrastructure for real-time markets.

Impact: 99.9% uptime • <50ms latency • 50+ concurrent streams

Open to Summer 2026
View Case Studies →Download Resume
SE

“Aarav shipped production-quality code faster than engineers with twice his experience.”

Senior Software EngineerTradingView project reviewer
TL

“The ZK implementation was clean and well-documented. He explained complex cryptography clearly.”

Tech LeadHashPledge hackathon judge
AJ

About Me

FinTech systems engineer — low-latency infrastructure, real-time data at scale.

My paper trading app crashed every time the market spiked. That frustration became my focus:systems that don't fail when they matter.

📍 Blacksburg, VA🎓 Virginia Tech

Case Studies

Problem → Approach → Impact

TradingView Recreation screenshot

TradingView Recreation

Cut chart latency from 200ms → 50ms, letting traders execute before competitors.

Solo build • 3 months • Replaced vendor tool

Problem

Trading platforms lag during volatility—exactly when speed matters most.

Impact

  • <50ms render latency (4x faster)
  • 50+ concurrent WebSocket streams
  • 99.9% uptime over 3 months

Approach

Canvas engine (SVG couldn't handle 50k+ points), WebSocket multiplexer, plugin system for indicators.

Lesson

V1 failed from premature optimization. V2 started with profiling.

HashPledge screenshot

HashPledge

Reduced verification from 5 days → 30 seconds with privacy compliance.

Team of 3 • Hackathon → Production

Problem

Manual document review hurt students. Privacy regulations blocked database solutions.

Impact

  • 5 days → 30 seconds
  • 200+ scholarships disbursed
  • Zero privacy complaints

Approach

Zero-knowledge proofs verify eligibility without exposing data. secp256k1 signatures for audit trail.

Lesson

Built ZK from scratch initially—switched to battle-tested library, shipped 4x faster.

ML Competition Pipeline screenshot

ML Competition Pipeline

Top 5% finish with zero manual tuning after setup.

Solo • 6-week competition

Problem

Manual calibration is slow (20+ min per tweak) and non-reproducible.

Impact

  • Top 5% finish
  • 40+ automated submissions
  • 10x faster iteration

Approach

Monotone transforms, surrogate model predicts LB score, safety gates filter bad candidates.

Lesson

The surrogate model was the breakthrough—predicting LB locally unlocked rapid iteration.

Interested? Let's talk →

Skills

Core Focus

  • TypeScript/JavaScript4yrProduction trading systems
  • Python4yrML pipelines, data eng
  • PostgreSQL3yrHigh-throughput data

Systems

  • WebSockets2yrReal-time streaming
  • Redis2yrCaching, pub/sub
  • Docker3yrDeployments

Frontend

  • React/Next.js3yrDashboards
  • Canvas API2yr60fps charts

ML

  • XGBoost/LightGBM2yrCompetition models
  • Pandas3yrFeature engineering
Download Full Resume

Let's Work Together

Open to Summer 2026 internships in trading systems, infrastructure, or ML.

aarav@vt.edu
GitHubLinkedInResume