18 Nov 2025·7 min read
Which LLM For Which Task (And Why I Didn't Self-Host)
Lessons from building a platform that needed to pick the right model for each task. Spoiler: 'just use GPT-4' isn't a strategy.
LLMOpenAIClaudeModel Selection+1
Notes
Things I've figured out
Writing helps me think. These are notes on problems I've solved, experiments I've run, and things I wish I knew earlier.
Showing 1-25 of 25 notes
5 Oct 2025·9 min read
Controlling 20,000 Requests Without Burning Money or Getting Banned
What I learned building request control for a multi-service LLM platform. Real patterns, real mistakes.
LLMRedisCeleryRabbitMQ+2
5 Sept 2025·6 min read
Why SEO is a Must-Learn Skill in 2025 (And It's Not What You Think)
SEO isn't just about rankings anymore. It's about visibility in AI Overviews, building trust signals, and understanding how machines interpret your content. Here's why every digital professional needs SEO now.
SEODigital MarketingCareerAI+1
12 Aug 2025·7 min read
What I Learned from SEO: Surviving the 2024-2025 Algorithm Chaos
The March 2024 core update changed everything. Here's what I learned about surviving Google's algorithm changes, why AI content strategies failed, and the lessons that shaped modern SEO.
SEOGoogle AlgorithmAI ContentE-E-A-T+1
25 Apr 2025·8 min read
When Matrices Stopped Being Scary (Linear Algebra for AI, Part 3)
How I learned matrices, eigenvalues, and SVD by connecting them to neural networks, PCA, and LoRA fine-tuning. The practical understanding that made AI systems click.
AIMathLearningMatrices+2
18 Mar 2025·7 min read
The Week Vectors Finally Made Sense (Linear Algebra for AI, Part 2)
How I learned to see vectors, dot products, and norms as practical tools for AI systems. The visual explanations that finally made embeddings and similarity click.
AIMathLearningVectors+1
10 Feb 2025·3 min read
Why I'm Learning Linear Algebra as a Web Developer (Part 1)
The mathematical foundations behind modern AI systems and why understanding them matters for building better applications. Part 1 of my learning journey.
AIMathLearningLinear Algebra
5 Dec 2024·8 min read
Why Did My Celery Workers Keep Dying at 3am? (Debugging Python Memory Leaks)
A deep dive into debugging Celery worker crashes in production. How I fixed memory fragmentation and database connection leaks in Django/Python using max-tasks-per-child.
PythonCeleryDebuggingProduction+2
20 Nov 2024·10 min read
How Did One Failed Request Turn Into 3,000? (Understanding Retry Storms)
A practical guide to preventing retry storms in distributed systems. Learn exponential backoff, circuit breakers, and jitter strategies that protect your services from cascading failures.
Distributed SystemsPythonReliabilityArchitecture+1
15 Oct 2024·8 min read
How Do You Stop an LLM from Inventing Prices? (Preventing AI Hallucinations in Production)
A practical guide to preventing LLM hallucinations in production systems. Learn how to validate AI-generated content using Pydantic schemas, fallback chains, and output validation before it reaches your customers.
LLMPythonAIPydantic+2
8 Sept 2024·9 min read
Why Was Our AI Taking 1.2 Seconds to Write an Email? (Optimizing LLM Validation)
A practical guide to reducing LLM response validation latency. Learn parallel validation, tiered checking, caching strategies, and streaming validation to cut validation time from 400ms to 20ms.
LLMPythonPerformancePydantic+2
22 Jun 2024·9 min read
Why Did Our Queue Crash on Black Friday? (Understanding Queue Sizing and Backpressure)
A practical guide to sizing message queues and implementing backpressure. Learn how to prevent queue overflow, handle traffic spikes, and build systems that degrade gracefully under load.
RabbitMQArchitectureCeleryQueues+2
10 May 2024·8 min read
Why Did Our Search Get Slower Every Month? (PostgreSQL Full-Text Search Limits)
A practical guide to understanding PostgreSQL full-text search limits. Learn when to stick with PostgreSQL and when to migrate to Elasticsearch, with real performance numbers from a 2M record vehicle search.
PostgreSQLElasticsearchDatabasePerformance+2
15 Mar 2024·9 min read
Why Were New Listings Taking 15 Minutes to Appear? (From Batch to Event-Driven)
How we reduced data processing latency from 15 minutes to 30 seconds by switching from cron jobs to event-driven architecture with AWS Lambda and SQS. A practical guide to real-time data pipelines.
AWSArchitectureLambdaEvent-Driven+2
20 Jan 2024·5 min read
The XLSX From Hell: When Shapes Break Everything
Auto-fill an Excel template with shapes? I tried 4 approaches. Only the dumbest one worked.
PythonExcelDataEngineeringAutomation
5 Jan 2024·6 min read
Building Dashboards Leadership Actually Uses
I built 5 dashboards. Leadership used 1. Here's what made the winner different and how I fixed the others.
GrafanaDataVisualizationMetricsProductThinking
15 Dec 2023·5 min read
Scheduled Jobs That Actually Recover From Failures
A script that runs isn't the same as a script that works. How I made nightly jobs self-healing with retries, checkpoints, and loud failures.
CeleryPythonAutomationErrorHandling+1
1 Dec 2023·5 min read
When the Frontend Sends a Query as a String
I was becoming a human SQL interface for 50+ engineers. So I built a query parser that let them self-serve with typo suggestions.
FastAPIPythonQueryParsingInternalTools
15 Nov 2023·5 min read
How I Turned a 4-Hour Report Into a Button Click
Watching an analyst spend 4 hours on VLOOKUPs every Monday, I built a Python automation that found 4x more anomalies in 45 seconds.
AutomationPythonPandasFastAPI+1
19 Aug 2023·9 min read
The Page That Made 147 Database Queries (Fixing N+1 in Django)
How I reduced an auction listing page from 147 queries to 3 using select_related, prefetch_related, and annotations.
DjangoORMPerformancePostgreSQL+1
12 Aug 2023·9 min read
Why Did Two Users Just Bid the Same Amount? (The Polling Problem)
When polling caused duplicate bids in our live auction system, we learned the hard way why WebSockets matter for real-time features.
WebSocketsDjangoRealTimeArchitecture+1
5 Aug 2023·10 min read
How Do You Process 200GB of Vehicle Data Every Day?
Building a daily ETL pipeline that downloads, deduplicates, transforms, and indexes millions of vehicle listings before users wake up.
ETLPythonAWSS3+2
29 Jul 2023·8 min read
How Do You Build a Search API with 30+ Optional Filters?
Configuration over code: how I avoided 30 if-statements and built a maintainable search API with Elasticsearch and Django REST Framework.
ElasticsearchAPIDjangoArchitecture+2
22 Jul 2023·7 min read
Why Did Toyota Searches Take 3x Longer Than Lamborghini?
Elasticsearch's default sharding spread our data randomly. Custom routing by vehicle make made searches 5x faster. Here's how.
ElasticsearchPerformanceArchitectureSearch+1
15 Jul 2023·7 min read
How Do You Deduplicate 3 Million Records Without Running Out of Memory?
When Pandas crashed processing 40 million rows, I discovered Miller CLI. Here's how streaming beats loading for massive CSV deduplication.
PythonETLDataPerformance+1