Hi, I'm Anuj.
I build reliable software
for real workloads.
Built backend systems, data pipelines, and ML workflows that handle real traffic and real data.
I build robust systems end-to-end -
simulation infrastructure,
ML experimentation workflows, and
web apps that ship cleanly. I care about
observability, debuggability, and reliability:
systems that surface their own failures so you
can fix them fast.
My background spans full-stack
development, data engineering, cloud infra on AWS,
and applied security research.
- Build SUMO-based traffic simulation infrastructure to compare traditional vs RL-based signal control across realistic city networks.
- Operate GPU-backed AWS EC2 instances for long-running training jobs with monitoring, logging, and auto-restart logic.
- Develop automation to run multi-seed experiments, aggregate metrics, and surface regressions in model performance.
- Create debugging workflows to identify silent failures in RL agents and environment configurations.
- Engineered automated pipelines processing 100k+ text and image artifacts on AWS GPU instances for fraud and threat analysis.
- Hardened ingestion with retry logic, rate-limit handling, and structured logging.
- Built LLM-powered tools to flag malformed inputs, reducing manual triage.
- Authored runbooks for pipeline recovery and reproducible experiments.
- Replaced manual data collection with automated pipelines, enabling reliable weekly ingestion of global political datasets for research analysis.
- Implemented retry logic, checkpointing, and logging to ensure reliable long-running data pipelines.
- Identified OWASP Top 10 vulnerabilities (XSS, IDOR, CSRF) in production systems — Netflix, Canva, Fitbit — using Nmap, jSQL, Burp Suite.
- Delivered repro-ready reports with impact analysis and remediation recommendations.
Remote execution platform to compile and run RISC-V code safely in the browser — with sandboxing, logging, and cleanup for isolated, debuggable runs.
Distributed calling system using an OpenAI-based agent inside AWS Connect flows, with structured logs, metrics, and retry/backoff for reliability.
Scrapes social media, extracts text from images via Vision API, enriches with LLMs. Validation + error categorisation + SQL persistence so bad records are isolated.
PowerShell wrapper that brings Linux-style chmod to Windows in a single command.
Coursework
DSA · Operating Systems · Systems Programming · Computer Networks ·
Big Data Programming · Software Development · Fundamentals of Data Science
Let's work
together.
Open to full-time roles, contract work, and interesting problems worth building. I reply fast.