Projects
Technical research, engineering implementations, and academic reviews.
Systems
Research & Academic
Open Source — vectools & n8n Workflows
A zero-dependency npm library for vector mathematics, plus AI automation templates published in the official n8n gallery — covering resume parsing and job search scoring with Google Gemini.
Audio-Visual Emotion Recognition In Continuous Domain
Master Thesis project implementing a high-performance framework for multi-modal emotion recognition using advanced self-supervised representation learning on audio and visual streams.
COLMAN: Collaborative Multi-Agent Navigation using Textual-Visual Embeddings
COLMAN explores object goal navigation using embodied AI agents, leveraging Transformer-based architectures and CLIP semantic embeddings for improved scene understanding.
IndoRE: Relation Extraction for Low Resource Indian Languages
The IndoRE project focused on relation extraction for three low-resource Indian languages: Bengali, Telugu, and Hindi using deep generative models and transformer architectures.
Deep Imitation Learning for Complex Manipulation Task from Virtual Reality Teleoperation
Research focused on robot skill acquisition through imitation learning and VR teleoperation, mapping raw pixels to complex robotic manipulation actions.
FAZE: Few-Shot Adaptive Gaze Estimation using Meta-Learning
Implementation of a personalized gaze estimation framework using DT-ED networks and meta-learning to achieve high accuracy with minimal calibration samples.
Systems
Research & Academic
Open Source — vectools & n8n Workflows
A zero-dependency npm library for vector mathematics, plus AI automation templates published in the official n8n gallery — covering resume parsing and job search scoring with Google Gemini.
Audio-Visual Emotion Recognition In Continuous Domain
Master Thesis project implementing a high-performance framework for multi-modal emotion recognition using advanced self-supervised representation learning on audio and visual streams.
COLMAN: Collaborative Multi-Agent Navigation using Textual-Visual Embeddings
COLMAN explores object goal navigation using embodied AI agents, leveraging Transformer-based architectures and CLIP semantic embeddings for improved scene understanding.
IndoRE: Relation Extraction for Low Resource Indian Languages
The IndoRE project focused on relation extraction for three low-resource Indian languages: Bengali, Telugu, and Hindi using deep generative models and transformer architectures.