Ajay Vikram P - Machine Learning Researcher

Selected Projects

SkillBlenderLC

Self-Evolving Meta-Skill Learning for Humanoid Robots

Language-conditioned extension of SkillBlender enabling autonomous skill discovery and reward learning for humanoids using LLMs.

ViTFuser

Global Context Understanding for Autonomous Vehicles

An attention-based deep learning architecture introducing multi-stage Vision Transformers and Feature Pyramid Networks to improve autonomous vehicle decision-making.

Ternarizing CNNs

Efficient Human Activity Recognition

Compressing a CNN model for human activity recognition using ternarization and quantization techniques for efficient deployment.

Hand Kinematics in BCI

State-Space Models (S4, LMU) for Decoding

Decoding continuous hand movements from brain activity using State-Space Models (S4, LMU) to predict 2D hand position.