Yinyin Liu

- Led the AI transformation of the Seismic platform, branded as Seismic Aura, delivering integrated, enterprise-grade agentic solutions that powered AI and Analytics capabilities and experiences across the platform. - Unified and scaled global AI and Analytics teams, driving strategic alignment, operational excellence, and talent expansion to accelerate innovation and execution. - Established and operationalized responsible AI governance in partnership with InfoSec and Legal, achieving ISO42001 certification and embedding robust safety, compliance, and monitoring frameworks - Delivered measurable business impact by scaling Aura Copilot to enterprise customers with emphasis of Quality, Performance, customer satisfaction and supporting revenue growth. - Fostered a culture of continuous AI-accelerated product discovery and delivery with product, engineering and design.

XOKind is a seed-stage AI startup. XOKind’s travel product, Una Travel, was featured as a top travel app by the New York Times, Apple “Hot this week”, CNBC, and USA Today. XOKind's AI platform exposed the multi-source ETL and the microservices-based AI capabilities behind Una Travel as a low-code B2B service. It leverages few-shot learning, AutoML and LLMs, to enable product teams to build, deploy and continuously evaluate and fine-tune their ML models. Use cases include: - Real time lead scoring, revenue/demand forecasting, tabular data prediction using AutoML - Semantic search and technical question-answering (Q&A) systems using LLMs - Explainable recommendations, personalized travel itinerary auto-completion - ETL as a data mesh connecting to a large set of unstructured and structured data sources and map to a canonical schema for downstream applications I was responsible for defining the technology roadmap, high-level architecture design, and data & AI strategy. I led the company’s product development team (ML, BE, FE and UX, design, contractors), and established a culture of continuous product testing with user testing platforms. Lots of hands-on work building as well using the technologies including: SW: Google cloud(GKE, bigquery, dataflow, GCR, Firebase, Pub/Sub), AWS(Sagemaker, custom labels), Fastlane, gRPC, Golang, Protobuf, Dart/Flutter, CI/CD with Github actions, Kubernetes, BloomRPC, Postman, Helm, Redis, Prometheus, Grafana AI: Haystack/Deepset, Huggingface LLM, PyTorch, Tensorflow, AutoGluon, SHAP, Pinecone, Elasticsearch, Spark, Beam Services: Stripe, Segment, Data analytics on MixPanel, Braze, OpenAI, Apple App store, Google play, neo4j, Snowflake Travel (e2e commercial-grade API integration): Google Places, Hotelbeds availability and booking, TripAdvisor, Yelp, BookingPal, Amadeus, Viator, GetYourGuide Product development: User testing platforms, continuous product discovery process, design sprint methodology, Figma, Framer

Established and led a global data science team (across US and Israel) and built capabilities in AI model performance optimization, NLP, RL, and model compression, etc. Defined a data science function that involved AI research, data science libraries and solutions, and algorithmic engineering for product engineering. Engaged with customers from CSPs and enterprises to apply AI in solving domain use cases. I was responsible for: - Setting the strategy of global data science function - Providing data science expertise to AI products development and delivering data science components for SW and HW products - Driving data science engagements with customers and partnering with business development and sales teams - Leading open-source project development and research efforts with a track record of releases and publications at top-tier AI conferences - Building external AI research and data science community relations that drove research collaborations with academia and adoption of open source libraries Our team built 3 open source libraries (~10k stars total): - RL Coach: https://github.com/NervanaSystems/coach (integrated with AWS Sagemaker RL) - NLP Architect: https://github.com/NervanaSystems/nlp-architect - NN Distiller: https://github.com/NervanaSystems/distiller I also led Intel's participation in Partnership on AI for 2 years and applied domain expertise to tackle issues in Fair, Transparent, and Accountable AI and AI for Social good. Research published in ICML, ACL, EMNLP. Work and opinions featured by VentureBeat and Forbes. Keynote speaker at O’Reilly AI conferences, GIMS, Girl Geek, and Women of Data Science. https://medium.com/@teamrework/top-30-women-in-ai-usa-edition-2f71ab177c72 https://www.bizjournals.com/sanjose/news/2018/08/29/women-artificial-intelligence-adbe-goog-vnda-intc.html https://aws.amazon.com/blogs/aws/amazon-sagemaker-rl-managed-reinforcement-learning-with-amazon-sagemaker/ (AWS Sagemaker adopted Intel RL Coach for their RL services)

Early developer of: - Deep Learning algorithms and frameworks - neon (https://github.com/NervanaSystems/neon): neon was one of the first widely-adopted DL frameworks with unbeatable full-stack performance optimization (before Tensorflow was released) - Deep Learning intermediate representation graph compiler and optimizer - nGraph (https://github.com/NervanaSystems/ngraph) Developed models and optimizations in NLP, Computer vision, object localization, image segmentation, etc. Drove algorithmic features for Nervana Deep Learning ASIC Led engagements with Nervana customers to apply DL in domains including Finance, Agriculture, Energy, Media, etc. https://www.investmentexecutive.com/news/industry-news/manulife-to-develop-ai-tool-thanks-to-new-partnership/ https://techcrunch.com/2016/08/09/intel-buys-deep-learning-startup-nervana-systems-for-a-reported-350-million/

Research, system and software engineering work for Qualcomm Zeroth project to incorporate neural processing into engineering solutions for mobile technology. Worked on the algorithms, system and software for: - Biologically-inspired vision systems using artificial or spiking neural networks - Adaptive system using neuro-modulators to maximize external rewards - Autonomous robotic systems with mapping and path planning capabilities - Spiking neural network simulation platform - Computer vision https://www.technologyreview.com/s/520211/qualcomm-to-build-neuro-inspired-chips/ https://www.technologyreview.com/s/531206/emtech-qualcomm-working-to-build-artificial-intelligence-into-smartphones/

Rapid prototyping and evaluation on various new technologies for potential new businesses - distributed robot control systems - computer vision - sensor technologies - mobile applications