TestMu AI (formerly LambdaTest)
Noida, IndiaMember of Technical Staff, Frontend
Nov 2025 to Present
- Built KaneAI's authoring cursor, a persistent insertion point in a step tree the LLM keeps rewriting: identity-anchored so it survives regenerated ids, self-heals across deletes, lives inside If/Else and While, drives Run From Here / Run Till Here and every insert. Shipped for both editors behind a flag; requested by 10+ enterprise customers.
- Built the If/Else and While block editor over LLM-generated steps: plain-English or operand/operator conditions, conditionals and modules nesting inside loops, blocks runnable as one unit, draft state coherent while the LLM regenerates any subtree. 90% of block-using tests nest.
- Replaced the step editor's full-tree save with a diff of changed rows: a typical long-test save went from 300 KB to 3 KB, across nested If/Else and While, behind feature flags with a full-payload fallback.
- Shipped bulk API import and response assertions for both editors: ~500 calls with assertions in one pass, roughly halving authoring time for API-heavy tests, with evaluation server-side so previews match what runs.
- Owned cross-project test-case sharing, taking test cases from locked-to-one-project to usable anywhere a user has access: permission-aware target picker, borrowed-case handling on every list and run, project-level settings. Found five silent data-loss bugs by checking every payload against the Go backend.
- Migrated 10+ components from @primer/react to the internal lt-components design system, drove WCAG 2.1 AA passes, and shipped the Accessibility Score widget.