I architect systems
that don't break.
Rust. C++. Zero-copy. I build the tools that other engineers depend on when it actually matters.
Most engineers build for the happy path.
I build for the moment everything goes wrong.
The Why
Production goes down. Everyone scrambles. Root cause? A hardcoded secret. A service that buckles at 2x load. I kept seeing it — teams shipping tools that passed demos but crumbled under real traffic. I decided I'd rather build the tools that survive the 3 AM incident than the ones that cause it.
The Path
Started with Python. Hit the ceiling fast — try scanning a million files when your runtime has a GIL. So I picked up C++, then Rust. Not from tutorials — from shipping real systems and debugging real failures. Offensive security taught me how things actually break. Algorithmic trading taught me that 50ms of latency is the difference between profit and loss. Every skill I picked up, I picked up because I needed it to solve a problem nothing else could.
The Mission
FlashAudit scans 847K files in under a second — Gitleaks takes minutes. My trading engine handles SEBI-compliant execution with a kill switch that works. My steganography tool hides encrypted payloads in images so well that steganalysis tools can't find them. I don't build proof-of-concepts. I build things that work under pressure, and I can prove it.
Here's what that looks like in practice.
The Arsenal.
Problem: A client's Gitleaks pre-commit hook took 4 minutes on their monorepo. Engineers disabled it. Secrets shipped.
Enterprise-grade secret scanner built in Rust with zero-copy I/O, memory-mapped files, and Rayon work-stealing across every core. Outputs SARIF so it drops into existing CI pipelines without rewiring dashboards. Built because the existing scanners couldn't keep up with real repo sizes.
Problem: A staging box kept getting probed with SQLi and path-traversal payloads. Userland proxies added 30ms. Native Win32 at the packet level was the only path under budget.
A multi-threaded HTTP proxy written in pure C — no frameworks, no dependencies. Inspects every packet for SQL injection and path traversal patterns, then auto-bans offending IPs. Built on raw Win32 threads because the problem didn't need abstractions, it needed speed.
Problem: Paste sites and combo lists leak credentials daily. Enterprise SOCs monitor for it; individuals get notified after the fraud hits.
A monetized personal SOC that monitors digital identities for credential leaks across dark web dumps and public repos. Integrated Stripe for billing from day one — this isn't a demo, it's a product with paying users and audit-grade PDF reporting.
DataCenter Manager
Private BetaConcurrent Resource Orchestrator
Problem: Every tutorial on Pthreads uses counters and toy queues. Real resource contention — thread pools, condition variables, lock ordering — only shows up when you build it.
Multi-threaded data center simulation built on raw Pthreads. Mutex-locked resource pools, condition variable signaling, and a custom thread pool. The kind of systems programming that doesn't show up in tutorials but runs under every cloud provider.
StegoVault
Private BetaSteganographic Encryption Tool
Problem: Encrypted blobs announce themselves on disk and in transit. Sometimes the threat model requires that the data isn't visibly there at all.
Hides AES-256 encrypted payloads inside ordinary images using LSB steganography. The output passes visual inspection and steganalysis tools. Built for scenarios where the existence of the data itself is sensitive — not just its contents.
NSE Trading Engine
Private BetaSEBI-Compliant Algorithmic Trading
Problem: Algo-Bot kept missing SEBI circuit-breaker timing and retail APIs throttled mid-session. A production rewrite was the only way to trade inside compliance windows.
Full-stack trading platform for NSE markets with a desktop-native Electron frontend and FastAPI backend. SEBI-compliant order execution, real-time P&L tracking, and a kill switch that actually works. This is the system that superseded Algo-Bot — same problem, production-grade solution.
GlassVault
Private BetaEncrypted Personal Finance
Problem: Every finance app I audited shipped plaintext SQLite under the hood, cloud-first by default. There's no personal-finance tool that trusts you to own your own data.
A personal finance tracker where every byte is encrypted at rest with SQLCipher. CRDT-based sync means your data stays consistent across devices without a central server seeing it. Built in Flutter because finance apps should run everywhere.
Algo-Bot
Private BetaEarly Trading Architecture
Problem: Manual entries kept missing the fills I was backtesting. Automating it surfaced every hole in the plan — this is where I learned what retail APIs can't do.
The first iteration of my trading engine — event-driven execution for NSE with automated risk-management kill switches. Taught me what retail APIs can and can't do, and directly led to the NSE Trading Engine rewrite with proper infrastructure.
Problem: In OR telemetry, deterioration is caught by eye — minutes after it starts. Monitors beep on threshold, not trend. That window is where patients are lost.
A fault-tolerant pipeline that processes heart rate and SpO2 data in real time, flagging clinical deterioration before a human would notice. Redis Streams for exactly-once delivery, Isolation Forest for anomaly detection. Zero data loss by design.
Staleness Scanner
Private BetaDependency Freshness Monitor
Problem: Every team I've worked with learned about a stale dependency when a CVE landed on a Friday. Dependabot spams; nobody triages. Staleness needs to be visible before it's urgent.
A full-stack tool that scans repos on a schedule, scores dependency freshness, and flags staleness before it becomes a security incident. Background workers via arq, PostgreSQL for persistence, Redis for caching. Built so engineering leads can see the health of every repo at a glance.
Problem: Every large-file transfer tool still round-trips through somebody else's server. You pay in upload time and in who-sees-your-file. Browsers ship WebRTC — we shouldn't need a middleman.
Direct browser-to-browser file transfer using WebRTC data channels. No upload, no server in the middle, no file size limits. The connection is peer-to-peer — your file goes from your machine to theirs, and nobody else sees it.
// Engineering Capabilities
peak throughput
files scanned in 0.4s
FlashAudit Core · enterprise monorepo
speedup delivered
vs. baseline tooling
FlashAudit Core vs. Gitleaks 8.18
minimum latency
p50 across hot paths
BioStream ML · 3K events/sec pipeline
What I Build For You.
Your Python service is slow? I've rewritten hot paths in Rust that went from minutes to milliseconds. I built FlashAudit this way — zero-copy I/O, memory-mapped files, parallel execution. The result was 10x faster than Gitleaks on the same workload.
I built FlashAudit because existing secret scanners were too slow for enterprise repos. I can do the same for your pipeline — scanning that runs in CI, catches credentials before they ship, and doesn't slow down your deploys.
I've shipped full-stack products end-to-end — trading platforms, SaaS tools, finance apps. Database to deploy, with auth, payments, and monitoring that works. I build for the engineer who inherits the codebase after me.
I don’t build software that works in demos. I build software that works at 3 AM when the on-call engineer is staring at a dashboard wondering what went wrong.
”