QA Engineer with 4 years of experience building automated test infrastructure for Web3 and SaaS products. Lately I build AI-powered QA agents — Claude tool-use pipelines that generate, execute, and self-heal end-to-end tests so humans don't have to babysit flaky suites.
- AI QA agents that turn session recordings into structured, executable Playwright test intents
- Self-healing locator engines — semantic (testId/role/label) selectors with confidence scoring, cached fallbacks, and Claude vision/DOM recovery when the UI drifts
- Claude tool-use agents (20+ tools) that drive Chrome through the accessibility tree, assert pass/fail explicitly, and stream results live over SSE
- Fail-closed verification — cryptographic checks (EIP-712 signature recovery) that validate what an agent actually did, independent of LLM judgment
- Upcover — Built the Dorian QA Agent: Jam recordings → Playwright tests via Claude, with a self-healing locator engine and real-time observability
- Polymarket — Built an AI-driven E2E QA platform for a Web3 prediction market, including a fail-closed CLOB order-interception pipeline
- Kinnect — Built QA Autopilot, an intent-first QA platform (FastAPI, Playwright, Next.js) with a 3-tier self-healing locator engine
- Brahma.fi — Sole QA Engineer; built the Web3 automation suite from scratch with Synpress, reaching 90% automated coverage and cutting manual effort by 60%
- NEAR Protocol — Resolved 100+ issues on BOS and MDAO, cutting user-reported bugs from 50% to 5%
- Comviva — Achieved 95% test coverage with 100+ Cypress automation scripts
- 200+ Web3 E2E automation scripts across Console.Brahma.fi, Swype.fun, and the Morpho AI Agent
- Multiple testing frameworks built from zero, containerized with Docker/Kubernetes and wired into CI/CD
- Deep experience with MetaMask/wallet flows and EIP-712 signed-order verification
- Shipped four AI-agent QA platforms in the last year
- 📧 Email: saswatmarpureddy@gmail.com
- 💼 LinkedIn: marpureddy-saswat
- 🌐 Portfolio: saswat-r-portfolio.vercel.app
- Testing frameworks and tooling for decentralized applications
- Contributions to the NEAR Protocol ecosystem
💡 Interested in agent reliability, intent-first testing, and making AI-native teams trustworthy. Open to collaborating on projects at the intersection of AI agents and QA — feel free to reach out!