~/notes
Notes · 3 essays
Long-form on MLOps, GenAI architecture, causal inference, and platform engineering — anchored to real production experience. RSS →
-
Causal Inference Over Deep Learning
mathWhen propensity matching and do-calculus beat a transformer — a FinOps lens on a 40% cost reduction at Amber Electric.
▸ Anchored to Amber Electric — 40% avg infrastructure cost reduction across 10k+ energy sites
- Causal Inference
- FinOps
- Time-Series
- Energy
-
First Principles Multi-Agent Orchestration
Why scaling-succotash is a distributed system first and an LLM stack second — Celery DLQs, circuit breakers, and the Kubernetes substrate that holds the whole thing up.
▸ Anchored to scaling-succotash — a production agentic search engine on K8s
- GenAI
- LangGraph
- Distributed Systems
- Kubernetes
-
Hidden Technical Debt in Machine Learning
mathWhy elegant theoretical data science requires massive infrastructure to scale safely — lessons from the Montu GCP ML Platform.
▸ Lessons from the Montu GCP ML Platform (67% ↓ lead times)
- MLOps
- Technical Debt
- Kubernetes
- Architecture