Writing

Technical Notes & Published Work

A curated index of writing around AI systems, model deployment, vision workloads, RAG, and accelerator-backed cloud infrastructure.

Personal Deep Dives

Coming soon: concise breakdowns of GPU inference bottlenecks, ML systems tradeoffs, and cloud-native deployment patterns from my own projects.

Focus Areas

GPU InferenceML SystemsCloud InfraRAGMultimodal

AWS Published Blogs

Selected external publications on applied AI workloads.

Meta SAM 2.1 in SageMaker JumpStart

Vision

Segmentation workflows and model access through JumpStart.

Llama 4 in SageMaker JumpStart

LLMs

Model family overview and deployment patterns.

RAG QnA with Llama 3

RAG

Retrieval, embeddings, and Llama-based answer generation.

Training Llama 2 with AWS Trainium

Training

Large-model training using Trainium on SageMaker.

Vision use cases with Llama 3.2

Multimodal

OCR, VQA, image reasoning, and multimodal workflows.

Advanced RAG patterns on SageMaker

RAG

Architecture patterns for scalable RAG systems.

Trainium and Inferentia with Neuron

Accelerators

Optimized Neuron environments for training and inference.