Nopporn Lekuthai

Nopporn Lekuthai

Dual-degree student in M.D. and Biomedical Engineering.Building high-impact NLP and production-grade ML systems.

Publications

EC-RAFT : Automated Generation of Clinical Trial Eligibility Criteria through Retrieval-Augmented Fine-Tuning

ACL 2025 findings
  • Integrated Retrieval-Augmented Fine-Tuning (RAFT) framework to generate structured clinical trial eligibility criteria that outperformed frontier models in a low-cost, efficient setup.

LAMAR : Clinically Aligned LLM-Generated Few-Shot Learning for EHR-Grounded Patient Question Answering

BioNLP-ST @ ACL 2025
  • First place in the leaderboard using a multi-agent pipeline requiring no fine-tuning, achieving a 20-point score increase over the baseline.

Experience

Biomedical and Data Lab, Mahidol University

Student Researcher

Aug 2024 - Present
  • Published two papers in ACL 2025: EC-RAFT and LAMAR.
  • Mentored junior researchers in advanced fine-tuning and in-context learning techniques for large language models.
PyTorch
HuggingFace Transformers
LLMs
RAFT

Agoda

Software Engineering Intern, GenAI

May 2024 - Mar 2025
  • Developed and maintained a Kotlin-based CV screening tool that automates the recruiting process, reducing manual screening by 1200± hours annually
  • Implemented a backoffice system for the CV screening tool, enabling HR to reduce onboarding time from 1 days to 15 mins
  • Reduced LLM costs by 4-fold by streamlining the model process, all while maintaining the same accuracy
  • Designed and implemented PoC for LLM-as-a-Judge and Generation Entropy system, which can be used to evaluate model hallucinations
  • Designed and Implemented a AI-assisted human labeling tool for better data quality and efficiency
  • Got a "Superstar Intern" Performance Evaluation and "Curiosity Catalyst" Award by the end of the summer internship
Kotlin
Gradle
Pytorch
Huggingface
MLFlow
Grafana
Docker
LLMs
React
Redux
Microsoft SQL
Spring

Eatway

Software Engineer, Co-founder (Technical)

Sep 2023 - April 2024
  • Create a first MVP using React Native (Expo) and Next.js for cross-platform experience targeting iOS, Android, and Web.
  • Employed Edge Functions and an Edge read replicas (Cloudflare D1) to minimize latency in server-client interactions in a read intensive environment.
  • Developed a Continuous Deployment (CD) pipeline using GitHub Actions for building and deploying native applications, web, and backend to EAS, Cloudflare Pages, and Cloudflare Workers, respectively.
  • Adopted a Monorepo strategy to enhance code reusability and efficiency.
  • Implement and host a self-hosted version of Pusher (Soketi) for a real-time and cost-effective communication between clients and merchants.
React Native
Hono
Cloudflare Workers
Tamagui
SQLite
Cloudflare R2
GitHub actions
Turborepo
Pusher

Thairath Online

Internship (Software Engineer)

May 2023 – Aug 2023
  • Finetuned a Language Model for a proof-of-concept autocompletion system to assist journalists and engineered a full-stack web application, deploying and integrating this model onto an EC2 instance using FastAPI and NGINX.
  • Created a React library along with authentication systems, widely implemented across company products, complete with thorough documentation.
  • Developed a trending words algorithm based on Longest Common Substring and finetuned Named Entity Recognition (NER) model.
  • Achieved top performance recognition in nearly every sprint (full-time employees included)
React
CSS
FastAPI
NGINX
Docker
Pytorch
Transformers
Pandas

Ramathibodi Poison Center

Contractor

Dec 2022 – Mar 2023
  • Created a web application that helps physicians calculate drug doses for pediatric patients.
  • Utilized Progressive Web Application (PWA) technology for a fast, network-independent, and installable web app in case of poor network conditions and emergencies.
  • Gathered and analyzed data for accurate ideal body weight estimation for Thai children from 0 months old to 19 years old.
React
Next.js
Tailwind CSS
Workbox

Ramathibodi Pitching Challenge

Co-founder, Head of Technologies

Dec 2020 – Jul 2021
  • An online research competition that helps more than 1,200 high school students from across Thailand to develop research skills.
  • Worked on web application for registration process and program landing website.
Python
Pandas
Next.js
React
CSS
Firebase
Cloudflare workers

Open Source

(selected)

t4-app

  • Developed an OAuth authentication process in React Native (Expo) that smoothly incorporates Supabase Auth, being the first Expo starter template that offers a budget-friendly auth provider alternative to the more widely used, but expensive, Clerk.
  • Fixed a couple of bugs in session management in the stack.

Projects

(selected)

Thai Sign Language Conversion

Electromyogram-based Word-level Recognition

  • Create a program to facilitate data collection and data tagging at the site of collection.
  • Develop an Attention-based Model that is capable of distinguishing between different sets of sign language gestures with more than 91% accuracy.
Pandas
Serial
Keras
TensorFlow

soadmai

  • Create a full-stack web application that enables students to find friends (and beyond) with common interests.
  • Achieved finalist status in a design thinking contest.
React
Next.js
Tailwind CSS
tRPC
PostgreSQL
Auth.js

Achievements

(selected)

HK Tech Southeast Asia Start-up Competition

Oct 2023 - Present
  • Currently in the ongoing competition, having advanced to the final round.
  • Created a Minimum Viable Product (MVP) web portal, encompassing UX/UI design, following the proposal and research of a project focused on cancer screening using exosomes.

SKILLKAMP by K-Bank and ascend group

May 2023
  • Achieved finalist status in a web development competition by developing a frontend of a comprehensive ecommerce website with a robust performance optimization.

Bulbbet

Thinc 10 days

Nov 2023
  • Secured second runner-up in a university-wide web development competition by developing a full-stack prediction market platform using best practices and the latest technology in web development.
  • Received the chance to mentor and guide five team members who were beginners in web development, providing them with comprehensive support to enhance their skill.