10university: "SRM Institute of Science & Technology",
11
12competitive: {
13codejam: "Rank 902 (90,000+ participants)",
14kickstart: "Rank 2905",
15codechef: "Rank 275 — Lunchtime Division 2",
16} ,
17
18/**
19 * Backend & AI Systems Engineer with 4+ years of experience building
20 * production multi-agent LLM systems, distributed backends, and SaaS platforms.
21 * Led end-to-end development from MVPs to production — including an
22 * AI-driven interview platform with LLM agents and LiveKit.
23 */
24
25} ;
26
27exportdefaultshikhar;
// profilepic2.png
EXPLORER/PORTFOLIO/experience.ts
$ git log --oneline --all
PythonNextJsTypeScriptReactJsGithub
+Built Case AI Auto Audit for Palo Alto Networks, scoring technical support engineers against 27 rules (SLA compliance, PII leakage, troubleshooting quality) using Gemini 2.5 Flash on Vertex AI.
+Automated ~90% of manual case reviews with daily batch pipelines, weekly per-engineer reports, and AI-generated improvement plans; processed 50,000+ real support cases.
+Validated LLM accuracy against a golden dataset; shipped end-to-end on a Python backend, Cloud Spanner, and a Next.js UI embedded in Salesforce LWC.
+Architected and delivered Metalynk, a B2B platform with 2,000+ active users, owning backend, schema design, frontend and AWS deployment (Next.js, NestJS, PostgreSQL).
+Led backend and AI workflows for Orterum, an AI conversational research platform running real-time voice agents (LiveKit, ~50 interviews/day) with LangGraph multi-agent orchestration and RAG; improved concurrent session stability by 25%.
+Built Tax Uttaram, Deloitte's internal taxation tool; solely designed its role-based access control and built streaming database-to-CSV exports.
+Orchestrated async pipelines on AWS ECS, S3, SQS, CloudWatch and EventBridge Scheduler; cut notification failures by 30% with retries and dead-letter queues.