Projects

A collection of AI/ML projects spanning computer vision, natural language processing, and systems engineering.

FeaturedNLP

LLM Fine-tuning Pipeline

Built a comprehensive pipeline for efficient LLM fine-tuning using LoRA adapters and 4-bit quantization. Supports multiple model architectures and includes automatic hyperparameter tuning.

10+
models
-70%
memory
3x faster
speed
LLMsPythonCUDALoRA
FeaturedComputer Vision

Neural Vision Transformer

Developed a novel attention mechanism that combines local and global context for improved image understanding. The model uses sparse attention patterns to reduce computational cost while maintaining accuracy.

94.5%
accuracy
86M
params
120 img/s
speed
PyTorchTransformersComputer VisionCUDA
MLOps

Distributed Training Framework

Custom distributed training framework supporting data and model parallelism for training large neural networks.

100+ GPUs
scaling
92%
efficiency
Auto-recovery
fault
PyTorchDistributedMLOpsNCCL
NLP

Multimodal RAG System

Retrieval-augmented generation combining text, images, and structured data for accurate question answering.

89%
accuracy
200ms
latency
1M+
docs
RAGVector DBEmbeddingsLangChain
Deep Learning

Neural Style Transfer App

Real-time neural style transfer application with custom styles and video support.

50+
styles
30
fps
iOS/Android
platforms
TensorFlowMobileComputer Vision
Computer Vision

Real-time Object Detection

Edge-optimized object detection system achieving 60fps on embedded devices. Uses TensorRT for inference optimization.

60
fps
45.2
mAP
16ms
latency
TensorRTYOLOEdge AIC++