AI Stack
Technologies I use to design and ship AI systems.
I group tools by capability instead of trends: model access, orchestration, serving, retrieval, evaluation, and platform operations. The goal is always the same — choose the simplest stack that can support robust experiments and reliable production behavior.
Models and LLM APIs
Choosing the right model family based on reasoning quality, latency, context window, and production constraints.
ML and experimentation frameworks
Core foundations for training-adjacent workflows, experimentation, and model understanding.
LLM orchestration and retrieval
Libraries and frameworks for agent graphs, retrieval pipelines, tool integrations, and structured prompting.
Serving and inference
Production pathways for APIs, self-hosted inference, throughput optimization, and low-latency delivery.
Data and vector infrastructure
Storage, indexing, and retrieval layers for knowledge systems and AI-backed applications.
MLOps, evaluation, and platform
Observability and platform capabilities that keep experiments reproducible and production systems measurable.
Cloud and workflow orchestration
Infrastructure patterns for event-driven AI systems, data movement, and production operations.