Jonesh Sharma

Architected a human-in-the-loop Claude Code workflow that migrated 11,000+ C++ tests from a proprietary framework to GoogleTest, reducing the full regression cycle from approximately six weeks to three days. Led a five-person team through the modernization, defining reusable message-driven integration and end-to-end test capabilities, including assertions, message-wait utilities, swappable environments, stubs, mocks, and an LBM transport layer. Enabled a self-organizing team through knowledge sharing and frequent proofs of concept, while maintaining architectural direction and demonstrating the solution to the head of engineering.

Implemented deployment quality gates in GitLab CI to prevent regressions from reaching production. Introduced pre-commit hooks that run linting, static analysis, and fast tests before code is pushed, catching issues early in the developer workflow. Led a test pyramid rebalancing effort, migrating slow-running mid-tier integration tests to fast, isolated unit tests and lightweight integration tests, cutting pipeline feedback time significantly while maintaining coverage. This shifted quality checks earlier in the pipeline, reduced flaky test failures, and gave engineers faster confidence before merging changes.

Built a lightweight utility around openapi-request-validator to automatically validate API requests and responses against an OpenAPI specification. The tool plugs into the automated test pipeline, intercepts calls, and flags contract mismatches between the API implementation and its published schema, so a payload that deviates from the expected shape fails fast rather than surfacing later as a production defect. It runs alongside existing functional API tests, so contract validation happens automatically without hand-written schema assertions for every endpoint. The approach caught several breaking schema changes before they reached downstream consumers, and reduced the manual effort needed to keep API contracts and implementation in sync.

Designed a shift-left mobile test framework that used an MITM proxy to intercept and mock network traffic, removing the need for time-consuming pre-test setup such as seeding accounts, provisioning back-end state, or waiting on external services. By injecting fixture responses at the network layer, Appium tests could start from a known state instantly, cutting suite setup time significantly and letting engineers run mobile UI tests earlier in local development instead of only in late-stage environments. The approach also made tests more deterministic by removing their dependency on shared, mutable back-end environments and reducing flakiness caused by test data collisions between parallel test runs.