M.S. in Computer Science @ UMass Amherst | Ex-Predoctoral Research Fellow @ IISc
I am a computer science student interested in Representation Learning, Generative Modeling, and Probabilistic Machine Learning.
| Category | Skills & Tools |
|---|---|
| 🧠 Core Theory | Generative Models (DDPM, VAE, GANs), Probabilistic ML (VI, EM, Bayesian Methods), Representation Learning, Optimization, Deep Metric Learning, XAI (GradCAM) |
| 💻 Frameworks | PyTorch, JAX, Scikit-Learn, HuggingFace Transformers |
| 🛠️ Languages | Python, C++, Java, SQL, Bash |
| ⚙️ Infrastructure | Git, Docker, Linux, Spark, Hadoop, TensorBoard |
| 📝 Specialized | LaTeX, FastAPI, Pandas, NumPy, OpenCV |
- Publication: IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2024)
- Developed nuclear norm–regularized autoencoders for low-rank latent spaces.
- Achieved FID (11.09 on MNIST) by fitting Gaussian Mixture Models (GMMs) in the latent space, demonstrating superior generative performance.
- Bachelors Thesis Report | GitHub
- Built a binary classifier for EEG signals by transforming data into wavelet scalograms and applying ResNet-18 transfer learning.
- Achieved 91.59% accuracy and utilized GradCAM to provide critical explainability for the medical diagnosis.
- GitHub
- Developed a multimodal model using cross-attention to merge features from wav2vec 2.0 (acoustic) and BERT (textual).
- Resulted in a 24% boost in F1-score for word-level stress detection over unimodal baselines, validating the multimodal approach.
-
Implemented core generative architectures, including Denoising Diffusion Probabilistic Models (DDPM) using a U-Net with spatial attention, and Vanilla GANs for data distribution mimicry.
-
(Other projects include work on Image Captioning (CNN+LSTM) and Siamese CNN for Change Detection.)
| Platform | Link |
|---|---|
| linkedin.com/in/rish01/ | |
| ML Blog | rish-01.github.io/blog/ — My notes on VAEs, MLE, and Probabilistic Models. |
rishab2001rs@gmail.com |