Hi, I'm Dina Taing
I am from Cambodia, a senior undergraduate in Lousiana State University Software Engineering, specializing in machine learning and high-performance computing. I have a background in research and currently develop and lead an open-source project for a research community.
Tech Stack
Experience
- Reduced staff search time by 70% by integrating hospital tools and resources into a unified MCP-driven platform.
- Delivered fast, reliable enterprise access by building and deploying an MCP server integrated with Microsoft Teams.
- Used Microsoft Foundry AI, Azure containers, and Redis caching to provide centralized high-performance knowledge access.
- Developed an open-source JetBrains standalone plugin using Java, Python, and Docker to simplify setup and eliminate system-level issues.
- Implemented Tobii-Fusion eye-tracker to detect developer gaze and distinguish between handwritten and AI/Copilot-generated code.
- Reduced configuration failure by 80% through optimization and improved developer experience.
- Reduced CFD computation time by 20× by optimizing C++ multithreaded execution.
- Integrated AI-based prediction models into the OpenFOAM workflow to accelerate simulation results.
- Hosted and organized hackathons for LSU students.
- Built and led development of the GDG@LSU.org website.
- Project Lead for GeauxApp, on track to launch for LSU students in Spring 2026.
- Ran workshops in C++, parallel programming, and Flutter.
- Increased HPX shared-thread execution performance by 30% by optimizing GCC-linked subsystem bottlenecks.
- Enabled higher scalability for parallel C++ applications.
- Developed a mobile app using Dart and Flutter to replace paper forms, improving construction site workflows.
- Worked on backend and Flutter app development for LSU students and research groups.
Highlighted Projects
Selected workA real-time hazardous infrastructure reporting app built with Flutter, AWS Amplify, DynamoDB, and AWS Lambda. Municipal empowers citizens to quickly report hazardous issues such as potholes, broken streetlights, flooding, unsafe intersections, and other municipal problems. Reports appear instantly on an interactive map, enabling rapid response and improved public safety.
Click to open full case study →
A cross-platform Pickleball tracking app built with Flutter and Supabase. This app helps players play with friends, create groups, track match results, and rank themselves to see who is the best.
Click to open full case study →
A student-built LSU campus discovery and event-sharing app that helps LSU students explore campus, discover events, follow organizations, and post updates. Designed, led, and coordinated by me with a 20+ developer team. Built with Flutter and Supabase, with a custom GCP VM middleware to optimize performance and reduce storage costs. Targeting a Spring 2025 release.
Click to open full case study →
An end-to-end mini NLP engine that builds a custom Byte Pair Encoding (BPE) tokenizer and an N-Gram language model from scratch. This project focuses on understanding how modern tokenization and classic statistical language modeling work under the hood, without relying on high-level libraries.
Click to open full case study →
A minimal but fully featured agentic controller loop that orchestrates an LLM with tools, JSON Schema validation, budgets, and retrieval-augmented generation. It demonstrates planning, argument repair, rolling summarization, and Chroma-based knowledge search on top of the OpenAI API.
Click to open full case study →
A practical LLM integration pipeline that wires together OpenAI (GPT-5-nano, GPT-5-mini) and OpenRouter (Qwen 3-8B) for large-scale question answering and automatic grading on SQuAD. The focus is on robust API integration, batch workflows, JSON-schema outputs, and end-to-end automation.
Click to open full case study →
A machine learning project that models umpire strike-zone decisions for LSU Softball using real pitch-tracking data. The system predicts the probability of a called strike based on pitch location, batter/pitcher handedness, and swing behavior. Built using neural-network transfer learning, model surgery, and custom visualization tools to generate detailed strike-zone heatmaps.
Click to open full case study →