Attention Residuals
Why every transformer ever built carries a structural flaw from 2015, and how the Kimi team fixed it by recognizing a pattern the field already solved once.
Practical engineering insights from the frontlines. Deep dives into data architecture, infrastructure economics, and system design.
Why every transformer ever built carries a structural flaw from 2015, and how the Kimi team fixed it by recognizing a pattern the field already solved once.
A practical framework for writing HLDs that actually remove uncertainty. Six phases, a review process, and a full walkthrough from problem definition to rollout.
The full interactive story — fifteen tools reshaping how we ingest, transform, stream, orchestrate, and serve data. A standalone microsite experience.
pg_textsearch brings BM25 ranking, 29+ languages, and blazing-fast full-text search to PostgreSQL. A deep dive into what this extension offers.
Every AI coding session starts from zero. Architecture decisions, progress, and hard-won lessons disappear between conversations. The Agentic Context System fixes this with a structured file approach that gives AI persistent memory across sessions.
A complete hands-on guide to Kubernetes covering 18 essential concepts. Follow along step-by-step on your own computer. Every command works, every example runs.
Streaming vs batch processing, modern data stacks, and architectural patterns.
Cost analysis, vendor comparisons, and build-vs-buy decisions.
Practical patterns, operational complexity, and real-world trade-offs.
I'm Boyan Balev, a software engineer focused on data infrastructure and distributed systems. This blog shares practical insights from building and operating data platforms at scale.
Every article comes from real experience in the trenches - no ivory tower architectures, just patterns that work in production.