I build machine learning systems with PyTorch and deploy them into Linux environments.
Most of my work revolves around ultrasound imaging, training pipelines, and efficient experimentation.

• 💼 Working at SmartAlpha
• 📚 Currently exploring: VQ-VAE, latent diffusion, small encoders, and making models behave more consistently
• 🧠 Interests: understanding how models learn, why they fail, and how to make them more consistent
• 🛠 I care about clean setups, clear experiments, and reproducible results.

🧰 Tech Stack

_{Tools I use daily or often experiment with.}

Languages

• Python • Julia • MATLAB • C • Rust (on-the-way)

Machine Learning

• PyTorch • PyTorch Lightning • TensorFlow
• NumPy • SciPy • scikit-learn
• Albumentations • Plotly

Data Engineering & Observability

• Apache Airflow for orchestrating reproducible ML and data pipelines
• OpenTelemetry (OTel) for structured logs, metrics, and traces across runs
• Grafana for monitoring pipelines, experiments, and system behavior
• Loki for centralized, queryable log storage

Models & Methods I use

• Real-time segmentation and classification models (MobileNet-based encoders)
• UNet and lightweight UNet variants
• VQ-VAE and latent autoencoders
• Denoising and diffusion-based models
• Learning dense, spatial feature representations from images

Infrastructure & Tools

• Docker • Kubernetes • Google Cloud • PostgreSQL • Helm
• Linux • shell scripting
• Git • MLflow/W&B-style tracking

📈 GitHub Stats

🗂 What I Work On Here

• Training and data pipelines for ultrasound models
• Experiments with generative and representation-focused methods
• Tools for data processing and evaluation
• Reproducible environments (configs, scripts, utilities, distributed training)

🔗 Find me online

_{Thanks for visiting ✨}