Michael V. Yudelson, Ph.D.
Email | myudelson.github.io | Pittsburgh, PA, USA
Professional Summary
Accomplished Data Scientist and ML/AI Engineer with a PhD in Information Science and a proven record of leading innovative analytical and machine learning projects producing quantifiable results. Seeking to apply my expertise in a managing role to make transformative contributions to organizational growth and success.
Key Achievements
- Advanced Computing and Machine Learning Development. Architected and implemented a high-impact BI and Gen AI solution at Chegg, Inc. that enhanced the efficiency of a critical solver component by 11% and directly contributed to product goals.
- Strategic Product Management. Spearheaded the development and deployment of a diagnostic model at ACT, Inc., which enabled a new data product that resulted in doubling client engagement with the company website.
- Research and Publishing Excellence. Co-author on 2 US patents, 90+ peer-reviewed papers; organized 5 workshops and 1 conference, delivered 20+ academic and industrial talks.
Professional Experience
Technology Lead
• Defining the product roadmap and dealing technological uncertainty.
• Developing value proposition and go-to-market strategy.
• Leading the full-stack development and architecture for an AI product.
Founder
• Designed and implemented a bespoke, scalable cloud infrastructure and ML pipeline, boosting the throughput of a social media publishing group 5-fold, and set up an asset management process for publication preparation.
• Built, deployed, and rigorously evaluated an assistive chatbot consultant, leveraging 1000 hours of expert-provided stream audio and leveraging an LLM-as-a-Judge methodology.
Staff Data Scientist, Learning Sciences and Data Science
• Spearheading efforts on building a pipeline for designing retrieval-augmented prompts.
• Created a process for LLM prompt evaluation and designed automatic diagram-rich reports.
• Designed and deployed a fine-tuned LLM for extracting math expressions from question text.
Principal ML/AI Engineer, AI and Data Science
Orchestrated efforts on ideation, design, and implementation of ML/AI-based product features offering automated answer-scoring capabilities as one of only two ML engineers in the company.
Senior Research Scientist/Senior Software Developer, ACTNext/Emerging Technologie
• Designed and built a diagnostic student learning model for the ACT Academy platform.
• Created an analytical pipeline for validating a suite of models.
Principal ML/AI Engineer, AI and Data Science
• Scaled a suite of affect detectors to serve up to 60 million students.
• Co-created a hybrid teacher-machine tutoring approach that performs automated hand-offs.
• Brokered acquisition of multiyear student data for theoretical and practical investigations.
Research Scientist
• Optimized a student model serving 0.5 million students a year to save them 37% of time.
• Developed and published an innovative approach to comparing learning models.
Post-doctoral Fellow, Human-Computer Interaction Institute
• Created a novel approach to modeling student knowledge acquisition and published the results in peer-reviewed journals and conference proceedings (this work was cited over 600 times).
• Developed an open-source tool for building Bayesian models at scale.
Education
Doctor of Philosophy in Information Science, Summa Cum Laude
Dissertation: Providing Service-Based Personalization In An Adaptive Hypermedia System.
Candidate of Science in Computer Aided Design, Summa Cum Laude
Dissertation: Development of models and methods for advanced online training of CAD operating personnel.
Systems Engineer in Computer Aided Design, Summa Cum Laude
Skills
Programming
Python, C/C++, Java, SQL, R, Matlab, Shell script, GNU AWK.
Solutions
AWS (RedShift, DynamoDB Sage Maker), Docker, Terraform, PyTorch, TensorFlow, Postman, CI/CD, CNN, RNN, JIRA, Confluence.
Methods
Statistical analysis, model evaluation, data science & modeling, NLP, deep learning, SaaS, audio/video signal processing, speech transcription, educational data mining, user modeling, A/B testing.
Soft skills
Adaptability, collaboration, communication, decision-making, emotional intelligence, leadership, problem-solving.
Notable Projects
Chegg, Inc. offers multi-turn chat support in their app. I built an evaluation pipeline for the LLM prompt-selecting component. I owned the labeling tasks, coordinated subject matter experts, ensured data privacy and security compliance throughout the process, and built the process to automate graphics-rich weekly accuracy reports.
Designed, tested, and deployed a custom fine-tuned Large Language Model to extract math expressions from question texts for Chegg’s problem-solving system that serves a custom deep learning model. The extraction rate improved by 11%.
This cross-team project focused on building a service complying with IMS Global Caliper standard. I led efforts on operationalizing a cutting-edge ML user model. Also, I headed data alignment and model adaptation for two external partners that had tangibly different data granularity and rate.
This multi-institution project, funded by the US Department of Defense, focused on a hybrid human-machine tutoring. I led the work of integrating the machine tutoring data streams with the human tutor logs of our partners. Additionally, I headed the efforts of automating and scaling multiple user behavior detectors.
This is a collaboration with the University of Pittsburgh, and the University of Helsinki. I led the ETL of student data covering 10 semesters, headed the analytical work, and co-advised doctoral students. I was the major contributing author on resulting peer-reviewed publications.
I initiated this cross-team work to create a pipeline for iterative improvement of cognitive modeling component in Carnegie Learning's Cognitive Tutor. Results allowed to save users 37% of their time. The project required handling of close to a billion data records and extensive simulations and replays of user behavior.
I developed this open-source tool for building innovative hidden Markov models. It was used in multiple peer-reviewed publications. HMM-scalable supports standard algorithms as well as hierarchical factors, regularization, and parameter multiplexing.
Michael V. Yudelson © 2026
Adapted from 960 grid