Yung Chi Daniel Cho
Verified Expert in Engineering
AI/ML Engineer and Developer
Toronto, Canada
Toptal member since September 1, 2026
Daniel (Yung Chi) is a senior AI and machine learning engineer with over seven years of experience in production ML systems, LLM evaluation, and ML infrastructure. His expertise spans LangChain, Spark, and PyTorch for the travel and manufacturing industries. At Expedia, Daniel (Yung Chi) led evaluation design for an agentic trip-planning product, defining a custom rubric with 30+ metrics.
Portfolio
Experience
- Python 3 - 10 years
- Deep Learning - 6 years
- Software Engineering - 6 years
- Machine Learning - 6 years
- LangGraph - 2 years
- Retrieval-augmented Generation (RAG) - 2 years
- LangChain - 2 years
- LangSmith - 1 year
Preferred Environment
Databricks, Docker, Kubernetes, Ray, Datadog, GitHub Actions, Amazon Web Services (AWS)
The most amazing...
...ML pipeline I have built processed up to two billion rows per day and contributed to a $34+ million gross profit uplift at Expedia.
Work Experience
Machine Learning Scientist III
Expedia
- Led evaluation design for a top-priority agentic trip-planning and recommendation product, defining a custom rubric with eight metric clusters and 30 metrics for recommendation trust, hard-constraint adherence, routing quality, and conversational UX.
- Set up team dashboard reporting to surface aggregate experiment scores with LangSmith trace drill-downs partnered with analytics and annotators to align LLM-as-judge scoring with human labels for internal dogfood and product sign-off release gate.
- Specified offline evaluation endpoint requirements to mimic user interactions and capture traces, tool calls, latency, prompt versions, LLM-judge scores, human labels, and failure modes in LangSmith datasets.
- Developed a LiveKit-based voice review proof of concept with prompt design, LangSmith evaluations, review-gap analysis, and automated evaluation to mimic UX testing for spoken feedback collection and follow-up question quality.
Machine Learning Engineer III
Expedia
- Productionized three end-to-end Spark/Scala ML pipelines for churn prevention and incentive decisioning across US and UK customer segments, contributing to a $34+ million gross profit uplift.
- Scaled ML training and batch inference over booking events, clickstream, loyalty, and customer behavior data, processing up to two billion rows per day in Spark and Iceberg-based workflows.
- Built configurable rollout architecture for multi-brand and multi-market launches across Expedia, Vrbo, Hotels.com, the UK, Canada, and international markets.
- Implemented observability, validation, and quality controls using Datadog, Collibra, automated tests, and production monitoring to improve the reliability of customer-facing ML outputs.
- Developed cost and volume guardrails for incentive recommendations, including automated safeguards to reduce financial risk when audience size or campaign spend exceeded expected thresholds.
- Productionized a daily PyTorch destination-ranker training workflow by migrating Databricks experimentation into production pipelines, integrating upstream Iceberg data, Feast feature hydration, inference workflows, and evaluation metrics.
- Tuned daily ranker training automation around data windows, Feast hydration, and GPU runtime constraints, with training completing in approximately two hours on one GPU over multimillion-row parquet-backed datasets.
- Scoped incentive optimization approaches using OR-Tools and multi-armed bandits to improve treatment assignment and decision-making under budget constraints.
Machine Learning Modeling Engineer
Corning
- Led two LangChain/RAG chatbot projects for manufacturing troubleshooting and technical sales, converting domain knowledge into interactive assistants that asked contextual follow-up questions, recommended products, and routed qualified leads.
- Built a DMAIC troubleshooting assistant that reduced methodology onboarding from a three-day training course to a 30-minute guided chat workflow for manufacturing engineers.
- Developed ML surrogate models and optimization workflows for glass/polymer manufacturing processes, including evolutionary optimization and simulation automation that reduced engineering workflow time from four months to two days.
- Built data and digital-twin workflows using Kafka, Python, HPC simulation outputs, and ML models; presented internal applied deep learning research at the Corning-Merck Data Science Symposium.
Data Science Fellow
SharpestMinds
- Developed a corrosion pipeline machine learning model to enable image processing for pipeline corrosion maintenance.
- Applied cross-entropy and multi-label binary classification with data augmentation using PyTorch and analyzed results using Python Pandas and scikit-learn. Achieved 94% accuracy on test images.
- Curated 8000 public images of pipelines and corrosion. Processed images using PIL in RGB.
Engineering Data Analyst, Pipeline Integrity
Enbridge Gas Distribution
- Improved safety and service quality; maintained Enbridge's competitive position and reduced environmental impact by implementing pipeline asset risk management software.
- Modeled and formatted pipelines, MAOP data, digs & repairs records, and collected geographical information of land use and right of way.
- Created a data pipeline for a 100k-row and 30-column dataset to predict the probability of gas service disruption for pipeline risk management and reliability engineering.
- Implemented stress analysis algorithm for dent and corrosion and validated risk algorithm for corrosion, erosion, third-party damage, and incorrect operations.
- Extracted, transformed, and loaded (ETL) data from CSV, Excel, and PDF using Python Pandas to SQL server and applied a corrosion model in a user-defined function, saving the company $80,000 in data service charges.
Experience
Churn Prediction and Intervention Incentive System
Themed Reranking Model for Destination
Education
Specialization in Data Science
Toronto Metropolitan University - Toronto
Master's Degree in Computational Engineering
University of Toronto - Toronto, Canada
Bachelor's Degree in Chemical Engineering
University of Waterloo - Waterloo, Canada
Certifications
Agentic AI Essentials; Federated Learning and MLOps
Georgia Institute of Technology
LLM, RAG, and Agentic AI Training
NVIDIA Deep Learning Institute
AWS Certified Machine Learning – Specialty
Amazon Web Services
Deep Learning Specialization
DeepLearning.AI
Applying Machine Learning to Engineering and Science
MIT xPRO
Skills
Libraries/APIs
PyTorch, TensorFlow, Spark ML, XGBoost, Pandas, ArcGIS
Tools
Apache Iceberg, Collibra, Google OR-Tools, Apache Airflow, TensorFlow Serving
Languages
Python 3, Python, Scala, Java, SQL
Frameworks
Apache Spark, Hadoop, Spark, Ray, LangGraph
Storage
Data Pipelines, Datadog, Apache Hive
Paradigms
DMAIC, Model Context Protocol (MCP), Automation
Platforms
LangSmith, LiveKit, Databricks, Docker, Kubernetes, Amazon Web Services (AWS), Langfuse, KServe
Other
Machine Learning, Software Engineering, Model Evaluation, Statistical Modeling, Stakeholder Management, Machine Learning Operations (MLOps), Model Deployment, Model Monitoring, Model Tuning, Personalization, Recommendation Systems, CI/CD Pipelines, Sentiment Analysis, Communication, Data Science, Data Engineering, Time Series Forecasting, Feature Engineering, Artificial Intelligence (AI), Generative Artificial Intelligence (GenAI), MLflow, Feast, LangChain, Retrieval-augmented Generation (RAG), Kafka, GitHub Actions, Hugging Face, FastAPI, Computational Science, Statistics, Control Theory, Computational Fluid Dynamics (CFD), Deep Learning, Amazon Machine Learning, Agentic AI Systems, Quantization, Natural Language Processing (NLP), Big Data, Analytics, Data Management, Model Development, Monitoring, Computer Vision, Quantitative Risk Modeling
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring