- Implement machine learning (ML) infrastructure, enable data scientists and ML engineers to experiment and train ML models in an accelerated and reliable fashion - Maintain and optimize ML infrastructure to ensure the latest features, capabilities, and hardware are available to the machine learning team in an issue-free environment- Implement in-house data labeling solutions to support fast iteration ML development through high-quality dataset curation- Design, develop, and maintain ML CI/CD pipelines to achieve training, testing, deployment, monitoring automation, and support ML operations at scale- Develop and deploy ML model training pipelines to support fast model iteration- Deploy ML models into staging and production environments, satisfying performance, latency, and throughput requirements- Monitor and manage the lifecycle of ML models in production through the implementation of model monitoring mechanism, evaluation pipeline, emergency handling procedure, traffic routing procedure, and ML on-call duty process- Respond to alerts and troubleshoot issues related to model deployment and ML infrastructure- Implement security best practices for machine learning systems, ensure compliance with data protection and privacy regulations- Manage MLOps costs effectively to ensure both development goals and budgetary constraints are met

Leading the central AI team at Stingray supporting all business functions on AI needs.● Successfully built the AI team by selecting and hiring high-performance, best-fit talents based on a 10-factor evaluation system that measures candidates' technical knowledge and soft skills (ML, DL, NLP, software engineering, MLOps, statistics & other, communication, conflict resolution, learning, likeability).● Identified, selected and prioritized 60+ ML use cases through stakeholder interviews and by using a systematic framework that assesses AI project ideas for strategic alignment, business impact, risk ,and technical difficulty.● Designed, created and deployed Intelligent Survey Engine that solves major survey pain points such as lack of empathy, redundant questions, and low quality responses using a deep learning driven, retrieval-based chatbot.● Built and deployed the multilingual version AI Categorizer based on BERT. The model supports 100+ languages.

● Led an AI team of 8 members and managed AI development end to end including data labeling (processes and infrastructure), ML modeling, and deployment. Achieved a 100% ML project deployment rate through rigorous project qualification, results-driven project management and data/evidence-based development.● Achieved 100% reproducibility in production models and 99%+ decrease in ML model deployment time (days to minutes), by standardizing deployment procedure, adopting DVC, GCP Prediction Routine and building a training and deployment pipeline.● Redesigned a topic system of 1,300+ topics by introducing the concept of industry grouping and topic-tag combination. The new system Topics 2.0 is infinitely expandable by adding new tags while providing details at an arbitrary level of granularity to support actionable insights generation. As a result of the new hierarchical design, topic overlapping, a major factor in suboptimal topic classification performance, is completely eliminated.● Designed, built, and deployed the Topic Miner app - a tool for the research and labeling team to automatically identify topics within a given body of text using unsupervised learning topic modeling techniques.● Led product management for the internal data labeling tool that enabled live monitoring of production models, and achieved human-in-the-loop model improvement through user flagging, random QC, and monthly retraining.

● Built AI Categorizer, a multilabel text classification DL model fulfilling 50% increase in topic prediction accuracy and 90% decrease in training time, by applying vector-based algorithms, improving data quality, leveraging parallel processing and creating a two-tier model system (topic model vs. tag and sentiment model).● Built Categorization Eligibility Classifier that identifies customer responses’ topic eligibility. Compared with the legacy manual process, the model captures 30% more eligible replies to populate the dashboard and increased identification precision by 240%.● Conceptualized, designed, built, and deployed Topic Performance Forecaster that allowed retailers to predict NPS scores based on topic performances, enabling live customer satisfaction monitoring.

As the mentor for Udacity's data science nanodegree programs including Machine Learning Engineer, Deep Learning, Data Analyst, Business Analytics and Data Foundations, I help Udacians to acquire data science skills using my unique capability to turn complicated concepts into easy-to-understand vivid examples.Coached thousands of Udacity’s data science students from around the world on subjects including data manipulation, data visualization, data analysis, machine learning, and statistics, helped students debug code and improve projects, received numerous student commendations thanks to my sound knowledge of various data science topics and effective communication skills.Aa career counsellor, I help Udacians unleash their potential by enabling them to transition to their dream roles. Helped hundreds of students define job search strategy, craft personal branding message and devise job search action plan. As a result, numerous students received job offers from iconic firms including Facebook, BMW, Amazon, Uber and many more!

- Designed and implemented accuracy tests for mobile device fingerprinting algorithms, using a systematic process including data acquisition/sampling, data wrangling, statistical modeling, statistical inference, data analysis and recommendation. Achieved high precision and low miss rate for fingerprints-based attribution algorithm;- Led and successfully completed the Tableau WoW Sales Report Project that transformed a heavily manual weekly reporting process into a streamlined dashboard based workflow, by engaging stakeholders, managing business requirements, securing data from various sources with data wrangling and data connection (QuickBooks, Salesforce, Postgres), designing dashboards and iterating based on stakeholder feedbacks. - Worked as a statistics subject matter expert for multiple projects, provided support on experiment design, sampling method, statistical inference and business analysis.

Working with large data sets, I leverage data wrangling, data mining, statistical modeling, machine learning and data visualization techniques to extract insights and make predictions, completed numerous data science projects from end to end.

Responsible for full cycle of consulting engagement management from business development, project management to implementation and client relationship management. Managing all aspects of assigned projects, including proposal development, project plan development, team management, data analysis, presentation development, and delivery of client presentations. Some achievement highlights are:✔ Developed a growth strategy for a dental care chain, involving expansion strategy, business growth opportunities, marketing strategy and performance management plan; ✔ Evaluated existing business operations for a financial services firm, developed a business plan to increase revenue;✔ Advised a social media start-up, covering market segmentation and competitive strategy;✔ Conducted a complete analysis and cost allocation for a language services agency, allowing the organization to make cost-based pricing decisions that ensure profitability;✔ Advised a mobile e-commerce start-up on its go-to-market strategy (business model, market segmentation, distribution channel and competitive analysis).

As a management consultant to the Executive Director, I worked on a variety of strategy and operations consulting projects. Some achievement highlights are:✔ Built a pricing model that allowed management to clearly understand the numerous service offerings’ price change impact on bottom-line; worked with Business Development to assess sales efforts effectiveness;✔ Successfully initiated, led and implemented the 1st internal service level agreement project in MCIS history.✔ Completed the Call Center Operations Innovation Project, identified 11 operational improvement initiatives that correspond to total potential cost savings/avoidance of $758K. Three top priority initiatives were implemented immediately.✔ Completed call center service level analysis. Based on a robust analysis of call handling demand on an hourly, daily, weekly and monthly basis, built a call center scheduling model that allows management to forecast call volumes and clearly understand call center service level's business impact, staffing requirement and optimal scheduling pattern.✔ Championed the Standard Operating Procedure (SOP) project, finalized and secured approval for the SOP based on existing processes, best practices and front-line and management input. Organized and facilitated employee training accordingly.