Ph.D. Researcher in Computer Science
Computer Vision · Multimodal AI · Robotics · Model Robustness
I’m Aaditya — a systems thinker at the intersection of research and engineering. My work bridges perception and intelligence — from robust vision transformers to retrieval-augmented solvers, and from gesture-based AMRs to scalable generative pipelines.
Currently pursuing a Ph.D. at UCF, I’ve previously engineered solutions at Bosch, Siemens Energy, and AlphaBake, and built models that aren’t just smart — they’re resilient, interpretable, and industrial-grade.
- Foundation Models in CV & NLP
- Zero-/Few-Shot Learning & Multimodal Inference
- Generative Modeling (Stable Diffusion, LoRA, IP Adapters)
- Robotics Perception & Planning (ROS, OpenCV, PID)
- Gradient-based Robustness & Adversarial Defense
| Project | Description | Stack |
|---|---|---|
ViTLoc |
Transformer-based Absolute Pose Regression using NeRF-style embeddings | PyTorch, Pytorch3D, Timm |
GEAR |
Gradient-based ensemble for GAN detection (CNN + ViT) with 97.75% accuracy | PyTorch, TMM |
RA-MATQA |
Retrieval-augmented Math QA using custom retriever & MWP-BERT | PyTorch, HuggingFace |
VOWEL |
Gesture-controlled mining robot with real-time vision feedback | ROS, LSTM, YOLO |
POBOT |
UAV-assisted UGV control pipeline using K-means + PID | OpenCV, ROS, Linux |
- Languages: Python, C++, MATLAB, Shell, JS, SQL
- Frameworks: PyTorch, TensorFlow, ROS, OpenCV
- Infra: Docker, GCP, Git, Weights & Biases
- Applied Theory: ViTs, LoRA, Retrieval-Augmented Inference, PID, Diffusion Models
- Website
- GitHub
- Email: [email protected]
finding creations between code and chaos