multimedia creator
“Meeting Niyathi during their interview for the Product Manager role at Microsoft was a standout moment in my career. They immediately distinguished themselves to all of us on the interview loop with their rare combination of passion, intelligence, creative problem-solving, and mental flexibility. Niyathi is more than just a skilled professional; they’re a dynamic force who leverages their diverse personal interests and professional experiences to drive innovative approaches & major impact. Their endless curiosity and embodiment of a learning mindset are truly inspiring.
In the Global org at Microsoft (2019-2020), Niyathi quickly made their mark. They developed expertise in complex domains like localization and machine translation, effectively driving highly technical projects to fruition. Their ability to navigate the nuances of cross-organizational collaboration with ease established them not just as a colleague, but as a go-to figure within the team. What truly sets Niyathi apart, though, is their genuine energy, enthusiasm, and the heartfelt commitment and care they extend to their teammates. I confidently recommend Niyathi for any team!”
Program Manager, Microsoft Translations4U / Machine Translation team
Machine learning for language localization (2019-2020)
Created, maintained, reduced costs and improved translation quality for Microsoft’s language machine translation (MT) ML models for Windows, Office, Xbox & Gaming, Teams, Exchange, Cloud + AI, and more
Saved 20% of our monthly translation costs (highest team cost, 15% of CVP org fiscal year 2020 goal for spend category) by investigating unusually large charges & making subscription changes; clarified pricing 36x higher than price assumed previously for months
Drove analysis which revealed that software tenants, our main investment up to this point, comprise only 1% of our actual translation volume for at least half of our tenants; planned multi-year work to bring remaining volume (99%) to our analytics and dashboards
Drove upgrade from statistical to neural and human parity MT models; drove support for language specificity (e.g. Portuguese - Portugal vs. Portuguese - Brazil), covering 100% of our M365 Commercial languages (99% of revenue for 2+ years, 99% of MAD for 6+ months) with neural MT models on a language subtype level; newly deployed models show no degradation and up to 35% translation quality improvement
D&I initiatives: trained teams and leadership on Disability Etiquette and Language Guidance after customizing CELA Accessibility training with the People in Office with Disabilities group (a new group I help lead) and team focus group; on org D+I and Onboarding v-teams