Engineering systems that remain reliable under uneven load and imperfect conditions.

• Built distributed data lifecycle systems processing 300TB per day and cut preprocessing time by 99 percent with automated purge pipelines.
• Reduced annual storage infrastructure cost by 2.2 million dollars through storage optimization and retention policy improvements.
• Designed Kubernetes-based orchestration and policy-driven workflows that improved uptime by 30 percent and reduced engineering overhead.
• Built a job runtime prediction system for distributed workloads with 10K+ dependent sub-jobs and delivered 3 percent P95 error for end-to-end completion estimates.
• Scaled a production API from 5 RPS to 300 RPS in three months.
• Architected hybrid FSS and object storage systems that reduced storage footprint from 3PB to 2PB using compression and dynamic tiering.
• Achieved 98 percent cost reduction in compliance pipelines using archival-tier storage and high-efficiency compression.
• Integrated AI-assisted developer tools such as Codex and Aider into internal review and merge workflows.
• Improved stability and observability for large-scale EMGC job execution and termination flows.

• Led a three-person team to build an MCP-like integration layer connecting LinkedIn, Amazon product listings, Gmail, and Outlook.
• Helped scale the platform from 0 to 10K+ users.
• Built a unified social publishing workflow across Instagram, Facebook, and LinkedIn.

• Processed and analyzed 10M+ job queue records across 80+ attributes to identify scheduling and resource allocation bottlenecks.
• Built a simulation platform using time-series and regression models for what-if analysis across OS, region, and hardware configurations.

URL Shortener Service

Backend service for unique and customizable short links with real-time redirection, expiration handling, and analytics.

Designed for low-latency access with a Redis caching layer and exposed through REST APIs.

JavaScript · MongoDB · Redis · Postman

Multilingual Information Retrieval

Deep learning project focused on semantic retrieval across multiple languages using transformer-based embeddings.

Built multilingual retrieval workflows using sentence-transformer style models for cross-language search relevance.

PyTorch · sentence-transformers