/built.
shipped. measured.

AI safety — scoped the jailbreak and prompt-injection eval rubrics, researched manipulation patterns specific to intent-based agents, and wired the mitigations directly into the pipeline so the user never sees them but an operator absolutely does.

Platform & tooling — drove infrastructure cost reductions, pushed workflow automation that cut manual deploy steps, and shipped an internal review agent that lints text + policy before any pipeline goes out (issues caught at draft time, not post-launch).

Customer & GTM — owned Intercom end-to-end (integration, behaviour-triggered automation, and the UX of every help conversation). Doubled as the product-support PM converting tickets into roadmap signal, and partnered with marketing on launch decks, blog narratives, and reproducible sales-demo scripts.

Started as Product Intern in Sep 2023. Followed agile practices — sprint planning, backlog grooming, and ongoing feature delivery alongside product and operations. The internship was a tour through the full product loop: research → spec → ship → measure.

Redesigned key steps in the student application journey to simplify decision flows and reduce friction. Introduced clearer progress-tracking features that improved application-step completion by ~9%. Ran user research with students and mentors to identify pain points in university selection and the broader application workflow — the discovery directly fed the next roadmap.

Promoted to Associate Product Manager in Jun 2024 after consistent ownership and cross-functional execution. Led agile sprints across product, engineering, and operations — owning friction-point discovery, feature iteration, and stakeholder alignment with the operations team. Application completion and engagement metrics moved ~15% during the APM tenure.

Product intern on the Bixby voice platform team. Owned regression-test workflows for new voice intents — wrote test plans, ran them across regional builds, and triaged the gaps before they reached production.

Sat between QA, the model team, and the platform PMs — translating engineering signal into product decisions about which intents were ready to ship and which needed another iteration. Every release-readiness call started from a shared dashboard I helped scope.

Where I learned that 'platform stability' isn't a single number — it's a portfolio of small disciplines that together decide whether 50M+ users trust the product the next time they speak.