Bellevue, Washington, United States
Highly accomplished and visionary Engineering Executive with over two decades of leadership experience in AI, Machine Learning, NLP and Analytics at top-tier technology companies including Google, Microsoft, IBM, and Bell Labs. Proven track record of defining and executing technical vision, strategy, and roadmaps for large-scale, multi-site initiatives that drive significant product innovation and business growth. Expert in building and scaling high-performing, agile teams, fostering a culture of engineering excellence, and delivering state-of-the-art GenAI, NLU and Data Analytics capabilities that empower millions of users and generate substantial impact. Recognized for pioneering contributions to AI/LLM and influencing industry standards through extensive publications, patents, and community leadership.
TALLIP publishes high quality original archival papers, survey papers, and technical notes in the areas of computation and processing of information in Asian languages, low-resource languages of Africa, Australasia, Oceania and the Americas, as well as related disciplines.
Thrive on creating groundbreaking AI solutions for challenging Google Search issues, and relish partnering with talented, diverse teams to deliver innovative, scalable products that enhance user experience and satisfaction. Recognized with Google Search Tech Impact Awards (2023, 2024) for GenAI features and the 2023 Google Tech Award for MUM, a flagship Search LLM initiative. Lead multi-site engineering, science, and data teams for Google's Search AI Platform, driving innovation in AI and LLMs: - Scalable GenAI Platform: Defined and led the Search GenAI Developer platform, democratizing GenAI use through robust APIs and multimodal experiences, improving devX and model performance. - Foundational Model Advancement: Co-initiated strategic efforts in advancing Gemini foundational models and Search-specific adaptations, focusing on encoder-based models, low-latency serving, and native RAG. - Agentic & Multimodal AI: Spearheaded agentic solutions, advanced RAG frameworks, and multimodal applications, showcasing Generative Search experiences. - Full-Lifecycle Model Development: Developed frameworks for optimizing model performance, enhancing personalization, recommendation, and discovery capabilities. Responsible for data creation/generation, model adaptation (SFT, LoRA, RL), and distillation.
Driving partnership with engineers, scientists and PMs across the globe to push the boundaries of AI and develop foundational technology that empowers developers to utilize large-scale AI in Search. Objective is to enhance the efficiency of large-scale AI models by reducing both computational requirements and the amount of supervision needed for them to adapt to new tasks. Leading engineering and science organizations responsible for knowledge graph augmentation and its applications to Multi-modal content understanding in Search. Area of investments include: - Content Understanding and Task Journeys programs leveraging NLU for Search. Established an LLM-based Knowledge Graph engine for advanced reasoning. - Advanced Search capabilities via cutting-edge NLU/LLM applications, transforming content understanding and delivering functionalities foundational to robust AI platforms. - AI Platform development integrating mined intents, entities, and facts to enable intelligent task completion, intent-based routing, and dynamic responses within multi-modal agent systems. - NLU: entity detection & linking, fact & relation extraction, semi-structured search, clustering, etc. - Content Understanding & Knowledge Engine Services - Knowledge Extraction and Representation - Complex Query Processing: leverage actions, entities and personal information to make this type of queries actionable: "show few lunch ideas with ingredients of this season that my daughter will enjoy”;
Provide executive advisory board support in the areas of AI/LLM/ML.
- Led the Cortana effort, developing and launching its Natural Language Understanding, Conversation Engine, and Dialog Management capabilities. - Built a high performance agile and coherent teams with focus on engineering excellence, science and innovation; this team also pioneered the development of the Harman Kardon Invoke smart speaker, powered by Cortana. - Drove the technical vision, strategy, and roadmap of the Cortana Cortex, impacting the overall strategy of Cortana; oversaw teams working on NLU, dialog systems, knowledge discovery, ranking, runtime, and quality measurements. - Managed and led multi-site teams of scientists and engineers in research and development of natural language understanding, dialog systems, search, and measurements utilizing machine learning, deep learning, reinforcement learning, etc. - Provided a dialog system platform enabling writing skills for digital assistants using procedural, event-driven, and example-based dialog systems. - Led the efforts working on runtime platform, dialog systems, ranking, parsing, spelling correction, knowledge-graph conflation, personal knowledge graph, and detection of domains, intents and slots. - Served as the bridge between business units, research and UX establishing milestones, delivering on them, enabling innovation and leveling it up.
Owned the Bing's core measurement program, ensuring that data and insights were at the heart of planning and prioritization: - Established the key performance indicators that defined success for the entire organization. Pioneered metrics for user satisfaction and abandonment that were adopted to set org-level goals, directly tying analytical work to business outcomes. - Led the analytics charter for user satisfaction, employing a broad toolkit of rigorous analytical approaches including A/B flighting and user behavior modeling to test hypotheses and drive product improvements. - Acted as the trusted strategic partner in critical launch decisions, building credibility by clearly communicating complex data-driven stories. Successfully influenced and convinced senior leaders from competing work-streams to align on a unified path forward. - Established metrics and a platform for measuring search and digital assistant quality using A/B testing and offline measurement. This involved leveraging large-scale data and distributed systems (e.g., Hadoop) to detect good/bad abandonment, model dwell time for click-level satisfaction, and assess novelty-based utility metrics and relevance dimensions in preference-based evaluation. Proposed metrics were used as ship criteria, promoting a data-driven organization. - Led a data-driven team whose software/algorithms ran on vast infrastructure, processing massive data volumes and impacting hundreds of millions of users daily.