Singapore
Hey! I'm Prannaya (pronounced prannay). I graduated from NUS High School of Math and Science back in 2023, and recently finished my 2 years of full-time national service as an AI Engineer and Researcher under the AETHER branch of RSAF's RAiD department, where I led the development of brain-computer interfaces for human performance monitoring and AI agents for automating the training process. Outside of my NS work, I previously served as an EXCO member at the tech-for-good charity, Better.sg from 2023 to 2025. In my time back in NUS High, I also served as the Overall-in-Charge of BuildingBloCS 2023 and the President of the NUS High CS Interest Group, AppVenture. Since 2020, I've found a special love for ML research, be it analysing IMU data to detect Parkinson's Freezing of Gait, or using Cuckoo Sandbox reports to identify malware samples. I previously developed models and libraries for LLM safety under Walled AI Labs, building an evaluation testbed named WalledEval.
Conducting applied research and building agentic solutions in areas spanning agentic AI optimization, agentic engineering harnesses, healthcare solutioning and end-to-end robotics.
Served my National Service obligations as an AI Researcher / Engineer at the RSAF Agile Innovation Digital (RAiD), under the Experimentations branch, AETHER. I primarily led an innovation team developing human-AI interfaces to monitor, enhance and augment controllers in the RSAF. In my time at AETHER, I: - Built, trained and evaluated Brain Foundation Models for generic psychophysiological understanding (via EEG, Eye Tracking, Speech etc) and prediction of cognitive states like Workload, Stress and Fatigue - Initiated and spearheaded development of Wingman, an experimental sandbox for Agentic AI tools and workflows within AETHER, including AERO (open-sourced at https://github.com/aether-raid/AERO) - Supported the development and evaluation of multi-agents for training, safety and operations within an integrated ATC Console - Supported the training of transcription models for ATC Communications (published in ICMCIS 2025) - Prototyped a MVP VSCode Extension for Automated Code Generation using simple LLM-based multi-agent systems, including an internal web search prototype (published in ICMCIS 2025, open-sourced at https://github.com/aether-raid/modular-search)
- Full-Stack development of an GPT-powered resume evaluator, led development of frontend with React.js - Developed several separate POCs and utilities for employees and clients in Temus
I worked with a bleeding-edge AI company under AI Singapore to develop a Benchmarking tool and a Proof-of-Concept News Classification application to pitch their potential models to Singaporean companies.
Project Title: Embodied AI for Computation Perception and Understanding of Spatial Designs Research Mentor: Prof Immanuel Koh We utilised transfer learning over large ResNet50 models over MIT CSAIL Vision's ADE20K dataset in order to classify HDB facades and interiors. This has been done as part of the Singapore University of Technology and Design (SUTD) Research Mentorship Programme (RMP).