Vision Intelligence Infrastructure: A high-performance, async control panel for human-in-the-loop AI annotation.

Decoupled ML microservices platform eliminating UI bottlenecks with instant rendering and async model workers.

• 🚀 Asynchronous ML: Non-blocking auto-labeling via PyTorch, LayerCAM & YOLO.
• ⚡ Zero-Latency UI: Hardware-accelerated React-Konva staging canvas.
• 🔄 Decoupled Scale: Distributed worker orchestration via Celery message brokers.

"Finding the needle in the haystack, from 400ft above."

A coarse-to-fine computer vision pipeline for efficient small-object detection in high-resolution (2K/4K) aerial imagery.

• 🎯 Smart Attention: Uses LayerCAM to localize semantic hotspots before inference.
• ⚡ 80%+ Background Skipped: Intelligently zooms into regions of interest, bypassing empty tiles.
• 🏆 High Efficiency: Outperforms blind sliding-window methods (SAHI) in speed and accuracy.

Transform any image into a masterpiece — in real-time.

Fast feed-forward neural style transfer generating stylized imagery in a single forward pass.

• 🚀 Real-Time Inference: Custom residual architecture trained with perceptual loss (VGG-16).
• 🔁 Artifact-Free: Instance Normalization for high-fidelity texture synthesis.
• 📦 Edge Ready: Full ONNX runtime export for edge deployment.

🧭 pygog (Google CLI Agent)
CLI for Google Workspace (Gmail, Drive, Calendar) with built-in natural language AI agent support.
<Python> <Google APIs> <LLM Agents>

📐 Depth Estimation + Semantic Seg.
Multi-modal depth completion (RGB + sparse depth + semantics) with NYU Depth v2 supervision.
<PyTorch> <NYU-Depth-v2> <Encoder-Decoder>