Research Associate at SUTD Singapore — working on world model learning, embodied AI, and Bayesian inverse RL on an MoE-funded Science of Learning grant.

I am an applied AI scientist with 7+ years across research labs and clinical AI environments in the US, Canada, and Singapore. My work spans embodied AI, probabilistic modeling, LLM post-training and inference, and XAI — grounded in computational neuroscience, cognitive science, and clinical data science.

📄 Website: hbk008.github.io

• World model estimation from kinematic data — Bayesian inverse RL over MDPs inferred from motion-capture sequences, grounded with MuJoCo humanoid simulation (SUTD, MoE SoL grant)
• Side projects — LLM architectures, post-training (RLHF, DPO, instruction tuning), efficient inference, and generative models

• Physical AI and world model learning for embodied agents
• LLM post-training, efficient inference, and AI infrastructure
• Mechanistic interpretability and XAI
• Computational neuroscience — EEG, motor control, brain-machine interfaces
• Clinical AI and precision health for neurodivergent populations

Languages Python (primary) · MATLAB · SQL (PostgreSQL) · Bash

ML / DL PyTorch · Hugging Face Transformers · Scikit-learn · NumPy / Pandas

LLM Training and Inference Familiar with: LoRA / QLoRA · mixed-precision training (bf16/fp8) · FlashAttention · ZeRO-stage distributed training · speculative decoding · PagedAttention / vLLM · RLHF · DPO · instruction tuning · quantization (int4/int8)

Generative Models VAEs · diffusion models (DDPM, latent diffusion) · autoregressive generation · vector quantization

Robotics and Simulation MuJoCo · Bayesian inverse RL · MDP / POMDP formulations · model-based RL

Biosignals and Neuroscience EEG / EEGLAB / MNE · cross-coherence and network connectivity analysis · wearable inertial sensors · Neural ODEs

Knowledge Representation Vector-Symbolic Architectures (HRR) · SNOMED-CT · FHIR

Clinical Infrastructure Azure ML · REDCap · POND network data pipelines

• Evaluating perceptual and semantic interpretability of saliency methods: A case study of melanoma — Applied AI Letters, 2022 · First author · DARPA/AFRL funded
• Digitized ADOS: Social interactions beyond the limits of the naked eye — Journal of Personalized Medicine, 2020 · First author
• Neural connectivity evolution during adaptive learning with and without proprioception — ACM MOCO, 2020 · First author ·
• Encoding medical ontologies with holographic reduced representations for transformers — ACM KDD Workshop, 2024 · Co-author
• A precision health approach to medication management in neurodivergence — medRxiv, 2025 · Co-author

Full list: Google Scholar

• MS Industrial & Systems Engineering — Rutgers University (thesis: deep learning for semiconductor process control)
• BTech Mechanical Engineering — Rajasthan Technical University