YASH RAJ · SINGAPORE
Computer Science student at NTU exploring useful software, AI systems, robotics, and the developer tools behind them.
01 / SELECTED WORK
Projects where I get to work through the interesting parts: model behaviour, systems boundaries, interfaces, and the awkward bits in between.
Open-source ML library
A native GDScript toolkit for learning agents inside Godot-built from the algorithms up, without external ML libraries.
Neural networks, CNNs, Q-learning, DQN, Minimax, evolutionary algorithms, ADAM / AMSGrad, experience replay, and target-network synchronisation.
Multi-agent career platform
A safety-conscious backend for onboarding, mentorship, feedback, and career development workflows.
Specialised LangChain agents use semantic retrieval and guardrails for PII redaction, prompt-injection detection, and bias filtering-exposed through containerised REST APIs.
BioHackathon · 3rd place
A machine-learning model for predicting cervical cancer from human RNA gene-expression data.
Built around a classification pipeline focused on a medically meaningful input space, with reported 96% accuracy and 94.5% recall in the project evaluation.
HackOSS · Technical Director
A pose-conditioned video-generation pipeline for stable, stylised stickman animation.
Leading model training and serving: curated normalised data, custom style and temporal LoRAs, a ComfyUI inference path, serverless GPU infrastructure, and cloud deployment.
ABOUT / WORKING NOTES
I'm Yash, a CS student at Nanyang Technological University. I'm drawn to technical work with both a conceptual and practical edge: teaching a game engine to learn, making agent workflows safer, serving models, or shaping a small tool until it feels right.
Right now, I'm especially interested in local AI, machine learning systems, robotics, backend architecture, and open-source developer tools.
More experiments on GitHub
YR / 202603 / EXPERIENCE
2025 - Present
Leading a seven-person team on Stickmanify's end-to-end video-generation system, from model training through GPU serving and cloud deployment.
2024 - 2025
Built a biometric authentication system and Django/PostgreSQL backend with role-aware workflows, scheduling logic, and REST API integration.
04 / TOOLBOX
05 / A FEW MILESTONES
06 / CONTACT