01
Evaluation, retrieval, and serving paths
How model-backed products depend on data quality, feedback loops, and operational discipline.
Explore notes ->
Field notes for AI, data, and BigData builders.
Exploring the systems, patterns, and tradeoffs behind production-grade data and AI platforms. From pipelines to models, from streaming to distributed jobs, these are the notes I wish I had when I was in the trenches.
Streamssignal flow
Jobsbatch + async
Latency12 ms
Throughput1.2M ev/s
Reliability99.95%
Writing from the machine room.
Deep dives, patterns, and real-world lessons from building systems that move, learn, and scale.
02
Trustworthy movement and shape
Streams, lakes, transformations, contracts, lineage, and the ergonomics that make data usable.
Explore notes ->
03
Distributed lessons that still matter
EMR, Spark, Kinesis, DynamoDB, Elasticsearch, and the reliability habits around them.
Explore notes ->
Latest notes.
Recent writing follows the practical overlap of AI systems, distributed processing, platform reliability, and clear operational tradeoffs.
View all notes ->
From EMR Escalations to AI Platform Reliability
A practical bridge from big data escalation work to AI platform reliability: the same habits show up in scheduling, observability, data quality, and recovery.
Observability Before Orchestration
Orchestration helps only after the system can explain itself. Metrics, traces, logs, and workload shape need to come before automation confidence.
SLO Thinking for Data Pipelines That Feed AI Systems
Data pipelines need reliability language too. Freshness, completeness, latency, correctness, and recoverability are better signals than green checkmarks.
The GPU Cluster Has a Data Problem First
The expensive part of an AI platform is not only the accelerator. It is the end-to-end path that keeps training and inference workloads fed, observable, and recoverable.
Follow the threads.
Curated paths through connected ideas.
