Blog

    Notes on engineering systems, architecture decisions, and practical scaling.

    LLMsAI ArchitectureEngineeringScalabilityNLP

    Beyond the Context Window: The Definitive Architect's Guide to Large-Scale LLM Summarization

    The context window is a lie. Even when your data fits, performance often doesn't. This guide explores Map-Reduce, Refine, Tree Summarization, and Semantic Clustering for production-grade AI.

    EngineeringAICloud Computing

    Beyond Tool Calling: Why Code Mode is the Final Boss of LLM Execution

    Traditional JSON-based tool calling is hitting a wall. Discover how Code Mode—pioneered by Cloudflare—is revolutionizing AI orchestration through sandboxed execution, massive token savings, and software-engineered agents.

    DatabasesScaling

    Horizontally Scaling PostgreSQL: A Practical Decision Framework (Without Sharding Too Early)

    Postgres scales reads easily. Write-scale is where complexity explodes. Here’s a clear framework: replicas + pooling, partitioning, sharding, Citus, or distributed SQL—with runbook-style guidance and failure modes.

    Distributed SystemsEvent-driven Architecture

    4 Kafka patterns you’ll actually use (and the trade-offs that matter)

    Mono-partition topics, key-based ordering, service-internal topics, and the outbox pattern—four real-world Kafka patterns and when each one is the right call.

    Documentation

    Architecture Decision Records (ADRs): the simplest way to stop repeating the same debates

    A practical guide to using ADRs to improve technical decision-making, alignment, and long-term speed—without heavyweight process.

    EngineeringDatabases

    SQL vs. NoSQL: Beyond Relational Data – A Deep Dive into Data Modeling and Access Paths

    A practical framework to choose between SQL and NoSQL by focusing on relationships, evolving query needs, and access-path design.

    Scaling

    Circuit Breakers in Distributed Systems: Fail Fast Without Cascading Failures

    A practical guide to circuit breakers: states, tuning knobs, metrics, and minimal runnable examples across Java, Node.js, Python, and .NET.

    Scaling

    Latency vs Throughput vs Bandwidth

    Why these metrics are related but not interchangeable, and how to optimize the right bottleneck.