Shift-left quality leadership, now architecting for agentic AI.
18+ years designing scalable test architectures and leading QA teams across enterprise SaaS and cloud — now driving AI-first quality initiatives, LLM evaluation, and agentic QA for production systems.
About
From shift-left automation to agentic AI architecture
I'm an AI Evaluation Architect and Agentic QA & Reliability leader with 18+ years designing scalable test architectures, driving AI-first quality initiatives, and leading QA teams across enterprise SaaS and cloud environments — spanning Oracle, Cisco (AppDynamics), Liberty Wireless Technologies, and Chargebee.
My hands-on practice runs from functional test automation framework design and test-early / shift-left strategy to one-click performance benchmarking and building Quality Centers of Excellence. Most recently at Chargebee, I built and led a 10-member performance test team from scratch and led that discipline into agentic AI — architecting Test Pilot (60% faster QA cycles) and ATPAR (performance testing cut from ~3 days to 1 hour). Over the last year I've extended this further into LLM and agent evaluation, and I teach it hands-on at SharadhaAI, the Agentic AI training program I run alongside my own projects.
Based in India. Open to Staff Engineer / QA Architect / Engineering Leadership roles in Agentic AI Evaluation & Quality, and to AI evaluation & governance consulting.
- B.E.Electrical & Electronics EngineeringBharathiar University
- AIPG Certificate, AI & MLIIT Guwahati
- ✓ISTQB — CTFL / CSTE CertifiedSoftware testing certification
- 🏆Devpost Hackathon — Agents for HumansAWS Strands Agents SDK, 2026
Experience
Where I've built quality systems
Four roles, one thread: shift-left test strategy, automation frameworks built from the ground up, and team leadership at scale — now extended to AI-native systems.
Built and led a 10-member Performance Test team from scratch — tool selection, shift-left CI/CD integration, and execution. Set up release gating for microservices and API performance benchmarking, and designed & rolled out an LLM-integrated testing platform. Architected Test Pilot (agentic requirement analysis, BDD test-case & automation code generation — 60% faster QA cycles) and ATPAR (performance testing cut from ~3 days to 1 hour). Partnered with Cloud Infra, SRE, GTM, and pre-sales on production RCAs and engineering excellence initiatives.
Led automation across 5 products with independent release cycles, owning test automation frameworks end-to-end. Designed and implemented full-cycle E2E automation from scratch, cutting release time from several hours to under 60 minutes. Built Python integration test apps and test harnesses on AWS EC2 and Kubernetes (EKS).
Led QE from requirement analysis through automated pipeline delivery and production support for a microservices architecture. Achieved build stability and functional integrity by identifying and automating P0 (BVT) tests.
Created and maintained the API functional test framework for Oracle's public cloud IaaS offering (OCI-C) using Java. Built the end-to-end BDD automation framework for the CGBU-BRM product using Java, Cucumber, and Selenium WebDriver — advocating test-early, shift-left testing, and establishing the org's Quality Center of Excellence.
Projects
Flagship work & hands-on builds
Frameworks built in production, plus independent projects built since exploring agentic AI full-time.
Test Pilot
An agentic AI toolkit that automates requirement analysis, BDD test-case generation, and automation code generation end-to-end — in production use at Chargebee.
↓ 60% QA cycle timeATPAR
A one-click performance test automation framework (JMeter + GitHub + JTL reporter) for continuous performance monitoring, built to make load testing a routine part of every release.
3 days → 1 hourAgentic Market Analyst
A multi-agent orchestrator (AWS Strands SDK + Groq, agent-as-tool pattern) for job-market & tooling-trend research. Caught a grounding/hallucination failure mode with an LLM-as-judge evaluation approach, then hardened it (rate limiting, prompt caching, structured output, steering) for Devpost's Agents for Humans hackathon.
Strands SDK + GroqEnterprise Agentic QA
A four-layer enterprise architecture for Agentic QA — a leadership-level blueprint for how organizations structure test strategy, agent evaluation, governance, and platform tooling as they adopt agentic AI.
4-layer architectureEnterprise Agentic Evaluation Platform
An enterprise-scale evaluation framework for agentic AI systems — deterministic and probabilistic testing, prompt-as-code versioning, security (OWASP LLM Top 10), and performance/chaos testing, unified in a single QA dashboard.
11-phase frameworkHealthCare RAG Assistant
A RAG-pipeline patient healthcare app, built during the IIT Guwahati AI/ML program, that helps doctors quickly assess a patient's medical history and plan effective treatment.
RAG pipelineSkills
QA architecture, leadership, and Agentic AI
Agentic AI & LLM Evaluation
- LLM output evaluation
- Prompt engineering & testing
- Agent testing frameworks
- RAG evaluation
Test Automation & Architecture
- Framework design
- BDD & requirement analysis
- Shift-left testing strategy
- CI/CD test integration
Performance Engineering
- JMeter benchmarking
- Perf test time: 3 days → 1 hour
- API test automation (OCI/BRM)
Leadership & Practice
- Built & led a 10-member QA team
- Quality Center of Excellence
- Cross-functional collaboration
- Technical mentoring & training
Tools & Technologies
The stack behind 18 years of QA — and the agentic AI toolkit today
From classic test automation to the agentic AI stack.
Also teaching
Founder, SharadhaAI
I run practical, cohort-based training on production agentic AI systems, RAG architectures, MCP, and AI evaluation frameworks — teaching the same discipline I bring to full-time engineering roles.