Building intelligent systems at the edge of data, decisions, and reality.
I design and build intelligent systems that operate under real-world constraints. My work converges at the intersection of machine learning, computer vision, and hardware–software co-design. I am driven by the challenge of creating decision-making systems that are not just theoretically sound, but practically indispensable.
- Lead Data Science Intern @ Blink Analytics | LLM development, RLHF pipelines, and evaluation.
- Advisor @ Mind Ripple | Leading a student-led intellectual forum.
Intelligence & Research
Vision & Robotics
Architecture & Interface
Systems & Infrastructure
A collection of long-horizon research projects exploring computer vision, interpretable AI, LLM security, and cyber-physical systems.
Integrated Footwear System & Computer Vision
A multimodal system for analyzing and scoring skateboarding performances using computer vision and gait analysis. By combining video-based trick detection with temporal segmentation, Talaria introduces objective evaluation into traditionally subjective sports judging.
Focus Areas:
- Action recognition & temporal segmentation
- Multimodal signal fusion (vision + motion)
- Objective performance scoring systems
🔗 Repository • 🌐 Website
Cervical Cancer Cell Classification
An end-to-end deep learning pipeline centered on model interpretability in medical imaging. The project leverages ConvNeXtV2 with concept-guided attention mechanisms to ensure AI predictions are transparent and clinically meaningful.
Focus Areas:
- Explainable AI (XAI)
- Medical image classification
- Concept-guided attention
🔗 Repository • 🌐 Website
Malicious Prompt Detection via LODO Analysis
A research-driven system addressing prompt-level vulnerabilities in Large Language Models. Introduces a Leave-One-Out (LODO) analysis framework to estimate token-level causal contributions, enabling intent-preserving prompt sanitization instead of binary blocking.
Focus Areas:
- Prompt injection detection
- Explainable NLP security
- Causal token attribution
Custom Drone Design & Flight Control
A ground-up autonomous drone project involving custom PCB design, sensor fusion, and low-level flight control. The system integrates an onboard microcontroller, IMU sensors, and ESCs, with a self-written flight controller for stabilization and control.
Focus Areas:
- Embedded systems & PCB design
- Sensor fusion (IMU, accelerometer, gyroscope)
- Control systems & autonomy
"I am the master of my fate, I am the captain of my soul." — William Ernest Henley
"People are neither rational nor irrational; they are simply human." — Daniel Kahneman