Loading...
Loading...
Let’s connect
Distributed systems, AI infrastructure, cloud architecture — or a problem that doesn’t fit any of those boxes. I’m open to internships, new grad roles, and good conversations.
About
I build AI systems efficient enough to actually ship — agent runtimes, LLM inference pipelines, and enterprise automation. MS CS at UIC on a 4.0, with three years across industry, a startup, and a research lab.
Path
Seven years, condensed — school, industry, research, and one registered company.
Develop and maintain AWS-based ETL and data-processing workflows that integrate data across university systems; build API-based ingestion pipelines and troubleshoot integration issues to keep production workflows reliable.
Built AI agent skills and LLM-powered workflows in ServiceNow. Cut processing time ~50% with a parallel-execution architecture and LLM token usage ~50% by refining context before model calls. Shipped MCP connector integrations for tool use and data retrieval.
Volunteer research on efficient LLM serving — semantic caching and KV-cache optimization. Analyzed KV-cache eviction and alternative attention papers, benchmarked inference approaches across latency, throughput, and scalability.
4.0 GPA, focused on making models cheaper, faster, and more controllable. Built polyserve (benchmarks LLM serving configs on your own GPU, then serves the winner), llmtrace (traces why vLLM requests are slow, down to GPU energy), Joule (budget-controlled AI agent runtime), and IWC (reproducible inference benchmarking) along the way.
Enterprise workflow and cloud-integration for Albertsons and Legend Biotech. Cut user-database sync time ~98% across local DBs, Azure, and ServiceNow; supported a 5-year $1M engagement. ServiceNow ITSM, CSM, HAM/ITAM, SAM delivery. Group Excellence Award.
Founded an MSME-registered D2C clothing brand and owned all engineering — React storefront with Razorpay checkout, plus automation for listings, content, order tracking, and marketing.
Selenium and REST API automation for CDSL's financial systems — 100+ test cases across trading workflows, identity sync, and Entra ID–ServiceNow synchronization.
Led a six-person team building a handwritten text recognition system (CNN + BiLSTM + CTC, 94% accuracy) and won at the national level against 50+ teams.
CGPA 8.48/10. GATE CS 2023, All India Rank 3410. Co-authored a research paper published in IJECERT.
Activity
Pulled live from GitHub. Some weeks are dense, some are not — that's honest.
Elsewhere
Straight from LinkedIn — current role and headline, rendered by their official badge so it never goes stale.
Toolkit
Grouped by where it sits in a system. Dots are a rough self-assessment; the graded version lives on the skills page.