MicrosoftSr Data & Applied Scientist
Nov. 2015Enterprise ML at scale: Architected end-to-end pipelines on Azure ML and automated deployments with Bicep + EV2, ensuring both compliance and agility. Data quality and anomaly detection: Built a drift and anomaly detection framework powered by statistical “fingerprints” of business data. By combining adaptive binning with tests such as Chi-Square, KS, PSI, and Cramér’s V, we enabled early anomaly detection and improved data reliability across pipelines. Partner enablement through AI: Created a certification recommendation model that analyzed candidate learning paths and suggested certifications aligned with the Microsoft Partner Competency Framework. This helped partners advance their Data & AI competency and improved specialization success rates with >80% precision. Operational efficiency with LLMs: Delivered an LLM-based incident categorization system that reduced noise by ~40%, automatically clustering issues into actionable categories and freeing up analyst bandwidth. Responsible AI and governance: Collaborated with Microsoft’s Central Governance Team to build a model registration catalog and embed Responsible AI guardrails — hallucination detection, jailbreak prevention, fairness, and transparency. This framework strengthened oversight and trust in GenAI systems. Data governance innovation: Combined advanced NLP and clustering techniques to create digital fingerprints of Azure Data Lake assets, making them easier to govern and discover across the enterprise. In parallel, developed a Data Lifecycle Management Framework that classified assets by compliance requirements, access frequency, and estimated carbon footprint. Automated tiering policies seamlessly shifted data between hot, cool, and cold storage, driving down storage costs, improving compliance, and supporting Microsoft’s sustainability commitments.