Retinal Disease Classifier
Designed zero-shot and linear-probe experiments with the FLAIR vision-language model for diabetic retinopathy and glaucoma detection, achieving 0.92 AUROC with only 5-20% labels.
View ProjectI build intelligent systems — from foundation models and LLM pipelines to production-ready AI tools.
Designed zero-shot and linear-probe experiments with the FLAIR vision-language model for diabetic retinopathy and glaucoma detection, achieving 0.92 AUROC with only 5-20% labels.
View Project
End-to-end pipeline that generates multi-strategy training data from documents, scores every example with rubric-based LLM-as-judge evaluation, detects benchmark contamination, fine-tunes a model locally with LoRA on Apple Silicon, and proves improvement with paired bootstrap significance testing. 184 tests, +16.8% ROUGE-1 gain.
View Project
Built a production-grade authentication service from scratch with JWT RS256, TOTP-based MFA, OAuth (Google/GitHub), refresh token rotation with theft detection, Redis rate limiting, and 136 automated tests.
View Project
I'm an AI / Machine Learning Engineer with a Master of Artificial Intelligence from Monash University. My work spans foundation models, large language models, and computer vision — from building retinal disease classifiers to fine-tuning LLMs for text-to-SQL generation.
Currently a Business Fellow at Perplexity AI, I prototype LLM agent workflows and explore AI product strategy. I'm passionate about taking projects from experimentation through to production-ready tools and demos.
Download ResumePerplexity AI
Selected into a competitive fellowship focused on AI product strategy and LLM-powered workflows. Prototyping no-code and low-code LLM agent workflows to automate internal analysis tasks.
Monash University
Designed experiments comparing zero-shot vs linear-probe adaptation of the FLAIR vision-language model for retinal disease detection. Built reproducible Python/PyTorch pipelines for the full research workflow.
1337 Ventures
Researched ~400 Malaysian startups, categorising companies by sector, stage and traction to map the early-stage ecosystem and support deal flow.
DeepLearning.AI · Coursera
DeepLearning.AI · Coursera
DeepLearning.AI · Coursera
Google · Coursera
Got a project in mind, a question, or just want to say hi? I'd love to hear from you.
pugalenthi0928@gmail.comOr send me a message directly