Rohit Agrawalstackclimb Rohit Agrawal
Principal engineer — AI platform, evaluation and release governance
- [email protected]
- GitHub
- Bengaluru, India · IST (UTC+5:30)
- Open to relocation worldwide, and to international travel
- Work authorisation: India · United States — H-1B approved
In short
Fourteen years deciding whether software was safe to release — and since April 2026, building AI systems that make that decision about themselves.
Quality engineering, test architecture and release governance across Oracle, Amazon, LimeRoad, Mobileum, Snapdeal and Subex, and most recently seven years at Oracle as a Principal Member of Technical Staff on Oracle Analytics.
The independent work applies the same discipline to AI: citation-grounded retrieval, disclosed uncertainty, explicit refusal, and evaluation gates that can block a release. Four of the six systems are gates.
Experience
Every figure in this section is self-reported. These are internal measures from the employers concerned and cannot be verified from outside them.
Principal Member of Technical Staff · Oracle
April 2019 — April 2026Oracle Analytics, on Oracle Cloud Infrastructure and Oracle Analytics Cloud — building enterprise quality engineering practice across the platform.
- Architected AI-driven quality engineering frameworks across enterprise cloud services — LLM-assisted test generation, intelligent test selection, self-healing automation and predictive quality analytics — reducing manual test design effort by 40%.
- Increased API automation coverage from 65% to 95%, reducing regression execution effort by 25% and production defects by 20%.
- Championed shift-right quality engineering with production monitoring, canary testing, synthetic monitoring and observability dashboards, reducing production incidents by 20% and accelerating root-cause analysis by 35%.
- Led the migration from Selenium to Playwright with Docker-based parallel execution, cutting regression execution time by 25% and flaky tests by 20%.
- Automated Terraform-based validation across 23+ Oracle Cloud regions.
- Defined enterprise-wide quality engineering standards and governance metrics while mentoring 11+ quality engineers and SDETs globally.
QA Engineer II · Amazon
May 2017 — September 2018End-to-end quality engineering for distributed enterprise platforms.
- Increased UI and API automation coverage by 25% and reduced regression execution time by 20%.
- Implemented shift-left testing across Agile and SAFe programmes, reducing defect leakage by 10%.
- Selected among the top five finalists for the Amazon D2AS innovation award.
Senior SDET · LimeRoad
October 2015 — May 2017Automation frameworks and payment integrity for a consumer marketplace.
- Built Selenium and Rest Assured automation frameworks, increasing coverage by 30% and cutting regression time by 25%.
- Validated payment gateway integrations across web, Android, iOS and mobile, reducing payment-related production defects by 25%.
Senior Software Test Engineer · Mobileum
April 2015 — September 2015Build sanity and big-data validation.
- Automated build sanity and Hadoop data validation, improving pipeline accuracy by 10% and reducing manual release-validation effort by roughly 35%.
QA Engineer II · Snapdeal
June 2014 — April 2015E-commerce platform quality.
Product Specialist · Subex
July 2011 — June 2014Telecom revenue-assurance products, working directly with customers.
- Multiple customer excellence awards for resolving critical customer issues, maintaining 99% SLA adherence.
- Chief People Officer recognition, and new business acquisition with StarHub.
Independent systems
Built since April 2026. Each row states the gate the system runs against itself — that is the through-line, and it is the same discipline as the fourteen years above.
CiteVyn
LiveCitation-grounded Q&A over official AI documentation. Answers quote their sources verbatim; where no source supports an answer it refuses rather than guessing.
Gate A 52-case golden run blocks index promotion. A single red case stops the release.
Quorum-AI
LiveOne question against four models in parallel, two rounds of mutual critique, then a synthesis returning consensus, disagreement, source support, uncertainty and a recommendation.
Gate Cost is approved before anything runs; any fallback or simulation is disclosed, never hidden.
SaafSaans
Deployed — sleeps when idleA Delhi-NCR air-quality companion scoring personal rather than city risk across 21 stations, with cited health guidance.
Gate Every reading is labelled live, deterministic fallback, or sample. The Hindi translation ships behind a banner stating no Hindi speaker has reviewed it.
NarraTwin AI
Phase 1 — No-GoGrounded walkthrough generation with citations, claim evaluation and consent checks.
Gate Its own release-readiness review reads No-Go, so it is not deployed. It is shown anyway, because the gate holding is the point.
EvalAxis
In progress — closedEvaluates LLM, RAG and agent changes against committed baselines — faithfulness, answer relevancy, hallucination, context precision and recall.
Gate Fails the CI build when quality regresses. This is the whole product.
Aegis Contracts
In progress — closedEarly-stage work on contract-shaped guarantees between AI systems.
Gate Too early to state one.
Technical
- AI quality engineering
- LLM and RAG evaluation · LLM-as-judge · hallucination detection · drift detection · AI observability · prompt and context engineering · multi-LLM orchestration · LangGraph · CrewAI
- Test engineering
- Test architecture · API, microservices and contract testing · performance · UI · database · defect prevention
- Automation
- Playwright · Selenium WebDriver · Rest Assured · Pytest · TestNG · JUnit · Postman
- Platform and DevOps
- Kubernetes · Terraform · Docker · Jenkins · CI/CD quality gates · AWS · Oracle Cloud Infrastructure
- Languages
- Java · Python · JavaScript · TypeScript · SQL · shell
- Data and backend
- PostgreSQL · pgvector · Redis · ClickHouse · MongoDB · DynamoDB · Oracle Database · FastAPI
- Observability
- OpenTelemetry · Prometheus · Grafana · Langfuse · Loki
Education and certification
- 2011
- Bachelor of Engineering, Information Science & Engineering — Visvesvaraya Technological University, India
- 2026
- Google Cloud Gen AI Academy APAC — Cohort 2 — Google Cloud
- 2026
- Generative AI Bootcamp — NSDC Skill India verified — GrowthSchool
- 2021
- Oracle Cloud Infrastructure Associate — Oracle
Recognition
- Oracle Rockstar Award (2024) — for API automation coverage and the drop in production issues
- Amazon D2AS Innovation Finalist — top five
- Customer Excellence Awards, Subex — multiple
- BugATAhon 2016 — regional 2nd runner-up