A \ V · Setup

What I Use

Hardware, software, tools, and models. The actual setup — no affiliate links, no sponsored picks. Just what earns its place on the desk.

🧠

AI & Models

Primary tooling
🤖
Secondary
OpenAI GPT-4o
Comparison baseline and multimodal tasks. Useful for evaluating where models diverge on the same prompts — disagreement between models is often the most interesting signal.
BenchmarkingMultimodal
🦙
Open Weight
Llama 3 / Mistral
On-premises deployments and sovereign AI requirements. When data cannot leave the building, open-weight models running on private infrastructure are not optional — they're the architecture.
On-PremisesPrivacy
☁️

Cloud & Infrastructure

Multi-cloud by principle
🟡
Primary Cloud
Google Cloud Platform
Primary cloud for AI/ML workloads. Vertex AI, BigQuery, and GKE form the backbone of most production pipelines. The data and ML tooling is second to none.
Vertex AIBigQueryGKE
🟠
Secondary Cloud
Amazon Web Services
SageMaker for model deployment where AWS is the incumbent. S3 as the universal data lake. Lambda for event-driven agent triggers. The broadest service catalog in the industry.
SageMakerS3Lambda
🔵
Enterprise Cloud
Microsoft Azure
Azure OpenAI Service for enterprise clients already in the Microsoft ecosystem. Active Directory integration and compliance tooling make it the path of least resistance for regulated industries.
Azure OpenAIEntra ID
🎩
Hybrid
Red Hat OpenShift
On-premises Kubernetes for healthcare clients with strict data residency requirements. The enterprise Kubernetes standard when the workload cannot live in a public cloud.
KubernetesOn-PremHIPAA
⚙️

Development

Daily drivers
🐍
Primary Language
Python
For AI/ML work, Python is still the lingua franca. FastAPI for agent APIs, LangGraph for orchestration, Pandas + Polars for data. The ecosystem is unmatched.
FastAPILangGraphPolars
🌐
Frontend
TypeScript + React
For AI-powered interfaces, React with TypeScript. The type safety catches the class of errors that are embarrassing to ship. shadcn/ui for components — the right abstraction level.
TypeScriptReactshadcn/ui
🔗
Protocol
Model Context Protocol
MCP is now the standard for tool integration in every agentic system I build. One server per integration surface. Composable, auditable, portable across any Claude-powered context.
MCPTool UseAnthropic
🕸️
Orchestration
LangGraph
Graph-based agent orchestration. Explicit state machines, human-in-the-loop checkpoints, cycle support. The right mental model for production agentic systems in regulated environments.
LangGraphMulti-Agent
📦
Containers
Docker + Kubernetes
Everything runs in containers. Docker for development parity, Kubernetes for production orchestration. Helm charts for deployment configuration. The baseline assumption for anything that needs to scale.
DockerKubernetesHelm
💻

Hardware

The physical layer
🖥️
Display
Ultra-wide Monitor
Horizontal real estate for terminal + editor + browser simultaneously. Once you work ultrawide, a standard 16:9 monitor feels like a port hole.
34"+1440p
⌨️
Input
Mechanical Keyboard
Tactile switches. The feedback loop between thought and typed character matters at the margin. Quieter than it sounds, better than membrane by a wide margin.
TactileMechanical
🎧
Audio
ANC Headphones
Active noise cancellation for deep work. The ability to create acoustic isolation anywhere — airport, office, road — is one of the highest-ROI tools I own.
ANCFocus
📚

Reading & Thinking

Off-screen tools
📖
Reading
Kindle + Physical Books
Kindle for travel and technical reading. Physical books for literature and anything I want to annotate properly. The marginalia habit is worth maintaining.
KindleMarginalia
✏️
Notes
Plain Text + Obsidian
Notes in plain Markdown. Obsidian for linking ideas across notes. The graph view occasionally reveals connections that weren't obvious. Everything is local, everything is portable.
MarkdownObsidianLocal
📡
Research
arXiv + IEEE Xplore
Primary sources for AI/ML research. arXiv for preprints the day they drop. IEEE Xplore for standards and established engineering literature. The peer-reviewed record matters.
arXivIEEEPapers
🗺️
Navigation
Paper Maps + GPS
Road trips: paper maps for planning, GPS for execution. The paper map forces you to understand the geography before you drive it. Old habit, worth keeping.
Road TripAnalog
Last updated March 2026 · Blue Planet · No affiliate links · All opinions are my own · Tools earn their place by being used daily