Michael Yang

- Led a high-performing AI team in delivering over 20 machine learning and generative AI projects, generating $1.5 million in revenue - Secured AWS's Generative AI Competency by successfully productionizing four strategic generative AI solutions across diverse business domains

- As the AIML Delivery Manager, successfully led the delivery of over 30 AIML projects. Collaborated with startups and medium-sized companies to implement cutting-edge AIML solutions, ensuring client satisfaction and project success. - Functioned as the primary Machine Learning Architect, contributing to the design and implementation of over 20 AIML projects. Applied AWS's ML well-architected framework pillars to develop end-to-end AIML systems across various stages of the ML lifecycle, ensuring robust and scalable solutions. - Implemented Gen AI applications such as medical SOAP notes generator, internal domain chatbots and image generator agent. Approaches used include Large Language Model (LLM)/Diffusion model fine tuning, Retrieval Augmented Generation (RAG), model quantization and low latency model serving on multiple GPU accelerators. - Drove the successful completion of AWS's ML Competency exam, earning Triumph Tech the highest recognition for ML capability as a consulting company. This achievement led to a 200% increase in ML sales opportunities from AWS for the organization, solidifying its position as a leader in the field. - Developed Proof of Concepts (POCs) such as cost effective serving of LLM and Diffusion Model to showcase Company's AIML capabilities, demonstrating innovative solutions to potential clients. - Managed and provided mentorship to a team of over 15 ML and data engineers. Through guidance and support, successfully facilitated the transition of 60% of these engineers into solution architects in their respective domains, nurturing talent and promoting professional growth.

- Led the AWS architecture design of the ticket master correlation engine “Dealer” that leverages 18+ ETL pipelines and 5 machine learning models (daily technician resource forecast, edge health score forecast, anomaly detection of network health, network outage detection and probability of network service affecting events) to submit proactive maintenance tickets to reduce over $6 million dollars of yearly transactional cost. - Designed and developed “Event Correlation” application to correlate and group redundant service tickets with new or working tickets via AWS EventBridge and Lambdas which reduced 23% of technician labor hours. - Developed MLOps framework with feature engineering, data preprocessing, data versioning, model training, model evaluation, model versioning, batch inference, data validation, data drift detection, target drift detection, model performance monitoring and automated notifications for model retraining. - Developed “Design of Experiments”; a scalable approach to design and execute AB testing by automating sample selection analysis (invariant metric check, sample size estimation, sample representativeness, bootstrap statistics) and monitoring of AB testing experiments. - Implemented NLP by leveraging discovered insight from call records and technician journal notes to network service affecting events which increased technician ticket actionability on average by 35% across the regions. - Migrated the application “Chronic” from on-premise to AWS by converting 40+ SQL scripts from Oracle to Presto and to Gremlin queries to enable utilization of Graph database.

• Collected, analyzed, and preprocessed raw operation data from various water and wastewater treatment facilities to develop forecasting models using multivariate regression for utility master plans. • Analyzed dataset through exploratory data analysis to define important features for blower energy consumption. • Developed a machine learning pipeline with Pyspark libraries using linear regression model to predict blower energy consumption based on air flow, operating pressure, temperature and inlet guide vane positions. • Developed a machine learning model to predict membrane permeability based on selection of operational features for the Spokane Water Treatment Plant which reduced O&M cost by 20% annually. • Performed sentimental analysis in customer feedback on Twitter in effort to identify trends between customer comments and effluent water quality for municipal water treatment plants. • Implemented ML model to SCREAM, Sewer Condition Risk-Enhanced Assessment Model, that predicts sewer pipe condition score which helped reduced 40% annually in pipe inspection cost for water and sewer municipalities. • Collaborated with global technologist and solution leaders on development of machine learning applications to analyze operation and asset management data in effort to optimize utility operation and maintenance cost for our clients.

• Collected, analyzed, and preprocessed raw operation data from various water and wastewater treatment facilities to develop forecasting models using multivariate regression for utility master plans.

• Assisted professor to plan class assessment materials • Evaluated/graded student assignments and created solutions for homework materials • Provided guidance to students during a filter design project and evaluated the design products

• Worked alongside with a graduate student and a professor to optimize parameters using machine learning for preparing iron nanoparticles used in environmental remediation applications • Designed and performed copper reduction and nitrate reduction experiments to determine reactivity of nZVI • Contributed as a co-author to the writing of a journal manuscript