Senior AI Developer
Phaphontee Yamchote
Professional Summary
I am a data and AI practitioner with a background that combines university teaching, machine learning research, and applied industry solution development. Over the past five years, I have worked across data analytics, ML research, and production-oriented AI systems, with the last two years focused more directly on real-world deployment and decision support use cases.
My core strength is math-driven problem solving: I prefer to understand the structure of a problem first, then choose the most appropriate method for the business context. I am particularly interested in work that connects rigorous thinking with practical implementation, especially in forecasting, recommendation, optimization, and data workflow design.
Core Strengths
Key Achievements
Faster Data Access
Redesigned a cloud-based data workflow for a retail fuel business, reducing report pagination time from around 5 minutes to about 10 seconds through better aggregation design.
Order Suggestion Workflow
Designed an ML-based order suggestion flow connected to full truck load optimization, helping bridge sales planning with downstream logistics decisions.
Adoption in Target Transactions
AI-assisted recommendations were used in approximately 80% of target transactions during February 2026, showing practical adoption by end users.
Research and Delivery Support
Mentored interns, junior developers, and graduate students in machine learning, ETL implementation, and research-oriented project execution.
Professional Journey
Career Evolution
Current
University Lecturer
Computer Science, Data Analytics, Machine Learning, and Mathematics
Teach technical and quantitative subjects while helping students build structured thinking and problem-solving skills. Also mentor project and research work across applied analytics and machine learning topics.
Parallel Work
Contract-Based ML / AI Solution Developer
Applied ML, Forecasting, Recommendation, Optimization
Work on real-world industry use cases involving machine learning pipelines, forecasting logic, recommendation systems, and business-facing decision support workflows.
Research Background
Machine Learning Research
Research, Experimentation, Model Design
Built experience through machine learning research and guided research implementation, especially in structured-data problems and mathematically motivated model design.
Foundation
Data and Analytical Work
From Analytical Work to Applied AI
Started from data-focused analytical work and gradually evolved toward data science, machine learning research, and applied AI solution development.
Trajectory
Career Evolution Summary
Data Analyst → Data Science / ML Research → Applied AI Solutions
Career progression has consistently moved toward solving more complex problems, connecting analytical depth with implementation, and designing solutions that can be used in real business contexts.
Featured Work
Order Suggestion and Full Truck Load Optimization
An applied AI decision-support workflow that connected machine learning-based order prediction with downstream logistics planning, helping improve sales support and transportation efficiency.
Business Problem
Sales teams faced a recurring challenge because turnover was relatively high, and new sales staff often lacked familiarity with customer ordering patterns. As a result, order planning depended heavily on individual judgment and prior experience. This created inconsistency in order suggestion and made downstream truck-loading decisions less efficient, especially when many smaller orders had to be grouped along shared delivery routes.
Context: The main challenge was not only prediction accuracy, but also how to support real sales decisions and improve route-based load planning.
My Role
I helped design the overall solution flow that connected ML-based demand or order suggestion with the full truck load optimization process. My role was not only to think about prediction, but also to make sure the output could be used in an operational workflow that sales and logistics teams could actually work with.
Key Responsibility: Making the solution useful in practice by aligning model outputs with business constraints, operational logic, and how experienced users actually make decisions.
Solution Approach
Prediction Layer:
Used machine learning to estimate suggested order quantities for the next cycle based on historical purchasing behavior and time-related patterns.
Optimization Layer:
Connected prediction outputs to a truck-loading and route-aware planning process so that smaller orders along similar routes could be grouped more effectively.
Feedback Loop:
Treated the system as decision support rather than pure automation, allowing recommendations to align with practical business judgment and operational constraints.
Deployment & Adoption
The solution was deployed as a batch inference workflow on Azure ML Studio, running once every two weeks. The output was used as operational guidance for planning rather than fully automated execution, making it easier for teams to adopt the system gradually.
Usage: The recommendation flow was used in practice and became part of the planning process for the target business group.
Impact & Why I'm Proud
~80%
AI-supported usage in target transactions during February 2026
Batch
Inference deployed on Azure ML Studio every two weeks
Reduced
Wasted truck capacity through better grouping of smaller orders
One of the most meaningful outcomes was that the AI-assisted recommendation flow was actually used in practice. In February 2026, approximately 80% of target transactions in the relevant group used AI-supported recommendations. Just as importantly, the outputs were considered reasonable by experienced sales staff, which increased trust in the system.
Why it matters to me: I am proud of this project because it was not just about building a model. It required connecting machine learning to a real business process, making the outputs usable for people with different levels of experience, and designing the workflow so that it created operational value rather than just predictive output.
I see myself as a practitioner who values rigorous thinking, real-world usefulness, and continuous learning. I am most motivated by work that connects analytical depth with deployable solutions and helps turn data or AI into something people can actually use to make better decisions.