MLOps, Deep Learning, Post-Training, and Embodied/Disembodied Agents.
- π¬ Manimator - Manim code generation agentic workflow
- π― MotionMarker - A lightweight motion data abstraction layer for Python animations
- π¬ Dialograph - POC: Framework for building proactive dialogue systems
- π₯ Animus - Agentic Blender-based animation system for autonomous scene and motion generation
- π₯ MomoPedia - Agentic platform for writing, reviewing, and approving articles about Momo around the world
- π€ Agentic AI Interview - AI native hiring agentic workflows for HR built using Google Meet SDK
- π PaperTrail - Tracking, comparing, and reasoning over AI research papers from arXiv
- π NetSecMLOps - MLOps infrastructure for network security models with automated training, validation, and deployment workflows
- βοΈ Makesense MLOps Pipeline - End-to-end MLOps pipeline with CI/CD, monitoring, and deployment automation
- π₯ PyTwerk - Minimal deep learning framework built on a micrograd-style autodiff engine for educational purposes
- π§© DeepBlocks - Block-based deep learning builder designed for teaching deep learning concepts
- π SwissKit - Modular Kaggle competition framework for rapid ML experimentation, competing and deployment
- π€ SPEECH - ASR fine-tuning codescript for Nepali speech-to-text using NVIDIA NeMo framework
- β±οΈ Timestamps & CTC Output - Audio timestamp extraction using CTC (Connectionist Temporal Classification) for speech recognition
- πΊοΈ Nepali Location with SymSpell - Fuzzy search system for Nepali location names using SymSpell spell correction algorithm
- β‘ pydplyr - Simplicity and conciseness of dplyr
- π Swin Transformer - From-scratch implementation of "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows"
- ποΈ Vision Transformer - PyTorch implementation of the original Vision Transformer (ViT) architecture
- π¬ GPT-2 from Scratch - Educational implementation of GPT-2 language model for CS224N final project
- π DSAwithPython β On progress.
- ποΈ Hanuman3D - 3D reconstruction of a Hanuman's Idol using Gaussian Splatting