I manage Dataminr’s end-to-end real-time AI platform (AI Ph.D. research, engineering, data, and data science teams) to build, deploy, and maintain AI at massive scale, ingesting and processing billions of data inputs per day (text, images, video, audio, sensor data) from over one million data sources for products used across all industries. - Scaled AI organization 7x, growing my management responsibilities to multiple teams over the years (Chief Scientist & SVP of AI until 3/24). - Increased LLM models in production (50x since ‘18), from hundreds of text-only sources in ‘18 to over a million in ‘24, expanding scope to include images, video, sensor, audio, and many languages.- Operational engineering responsibilities (24/7).

Developed and taught industry-focused course. Topics:Ethics & responsible AI, Transformers from scratch, Prompt Engineering, Fine-tuning of LLMs, RNNs, Stable Diffusion, Reinforcement-Learning from Human-Feedback (RLHF), connecting multiple LLMs (LangChain, AutoGPT) and deploying them at scale, Regularization, Pre-training, embeddings, and data augmentation techniques.Was also a Visiting Professor at:- KAIST (Daejon, South Korea, 2013): designed and taught "Human-Centered Data Analysis."- Universidad Carlos III (Madrid, Sept. '09-Feb '10): helped set up new graduate program in Human-Centered Computing (Data Mining, Social Network Analysis, Stream Mining, Interactive Media).

Managed engineering and science team. Revamped in-prod algorithms, data collection, & evaluation.

Led the company's AI/ML efforts. Launched first ML product, designed controlled experiments and set up the R&D strategy.

Managed the engineering team, and redesigned the flagship product and algorithms.

Started in Barcelona (Feb '10 until Dec '13), growing and managing teams of engineers and scientists from scratch in New York City, Sunnyvale, Bangalore, and Barcelona.Product impact highlights:o Delivered new user-facing feature across all Yahoo properties worldwide (5%-20% CTR increases, generating millions in revenue). o Delivered production algorithms on Yahoo search/homepage worldwide (1B+ users) for image classification, celebrity recognition, near-duplicate detection, image quality, video summarization, face recognition, and automatic thumbnail selection. o Created Yahoo Clues, a product to provide search insights and trending topics. Coordinated transfer of research from multiple teams and led interaction with Product, Design, Engineering teams, and SVPs.o Created various internal insights dashboards and reports (Flickr, News, Video) to impact business and product functionality (features, interface design, etc).o Launched US & Italian elections visualizations on Yahoo News.o Established an internal, global experimental innovation program, and obtained funding and led o Yahoo’s efforts in several EU projects on Data Driven Journalism.o Led design and implementation of advanced metrics (mouse tracking, social sharing) and A/B testing to produce insights (News, Video, Finance, Flickr, Image search). o Research and analysis of web-scale data to understand and model user behavior and engagement.

Established and managed the User Modeling and Machine Learning group, leading high-level meetings with strategy/business units to identify new data monetization and business opportunities.• Research and implementation of techniques for Social Network Analysis, and Visual Analytics applied to large-scale customer data. Results transferred and deployed in marketing campaigns.

Set vision for project on meeting interaction. Operational leadership, and planning of all scientific activities, as well as coordination of technology transfer efforts, and research in human-centered computing and interactive media.

Research on machine learning for video analysis and interaction, user studies, design, and implementation.

Research on statistical classification of videos using visual and textual features (e.g., WordNet).Internships:- IBM Tokyo Research (Summer 2001): Developed a supervised learning framework to generate personalized video digests from user input.- IBM Research Yorktown (Summer 2000): Studied quality corrections done by an expert, at the National Gallery of Art in Washington, DC, on digital images of art and developed an approach to semi-automate the correction process.- IBM Research Yorktown (Summer 1998): Implemented a new approach to automatically remove dithering screens from scanned documents, and created a high-end stereoscopic imaging framework for capturing valuable museum objects.

Research and implementation of machine-learning techniques for video indexing using audio.

Designed and implemented a GUI to monitor and manage AT&T’s Worldwide Intelligent Network.