Software Engineer at Equal Collectives (2025-05 – 2025-12)
Built and optimized backend systems for e-commerce review processing with focus on performance optimization and LLM integration
- Built and optimized the Lex backend handling 10k+ reviews/day, improving ingestion throughput by 94% (5s → 0.3s per 100 reviews) through API refactoring, bulk-write pipelines, and elimination of N+1 Prisma queries
- Designed a multi-stage scraping pipeline using Azure VMs, batching ASINs in groups of 3, implementing advanced filters to fetch significantly more than Amazon's default 100 reviews, and improving scraping reliability from 45% → 96%
- Implemented a BullMQ-based priority job system that orchestrates scraping, LLM evaluation, policy-violation detection, and reporting, reducing latency for high-priority tasks by 70%
- Integrated LiteLLM to centralize LLM usage, enabling seamless switching between OpenAI, Gemini, and Azure LLM models while reducing per-request latency and maintenance overhead
SDE Intern at Equal Collectives (2024-02 – 2025-04)
Improved web scraping infrastructure reliability and built scalable batch-processing workflows with serverless architecture
- Improved a mission-critical scraping pipeline's reliability from 45% to 96.5% by implementing anti-bot strategies, including Puppeteer Stealth, rotating residential proxies, dynamic delays, and exponential backoff retries
- Built an async batch-processing workflow that scaled to 100+ product audits per run, optimizing API usage and reducing redundant work with a 7-day caching layer (70% hit rate)
- Built a serverless PDF reporting system (AWS Lambda + Puppeteer) to generate customer-ready listing-quality reports with consistent formatting and low operating cost
