Things We've Learned
Articles from our team about data, AI, cloud—and the messy reality of getting them to work in production. No fluff, just what we wish someone had told us three years ago.
Why Your Agentic AI Project Will Probably Fail (And How to Fix It)
After watching three clients spend millions on autonomous AI systems that barely work, here's what actually matters: error handling, fallback strategies, and knowing when to just use a decision tree.
We Migrated 15TB to Snowflake. Here's What We'd Do Differently.
The good: query performance tripled. The bad: our first month's bill. The ugly: stored procedures don't translate automatically. Lessons learned from a six-month data warehouse migration.
You're Probably Overpaying for Cloud (A Lot)
Helped a client cut AWS costs by 63% without changing their architecture. The problem wasn't the cloud—it was Reserved Instances they weren't using, over-provisioned RDS instances, and S3 buckets no one remembered creating.
GPT-4 Is Not a Data Analyst (But It Can Help)
Real talk about LLMs in analytics: what works (data cleanup, SQL generation), what doesn't (complex statistical analysis), and why you still need actual data scientists.
Stop Saying 'Digital Transformation'
Your company doesn't need a transformation—it needs to fix its data quality, train its people, and actually use the tools it already bought. A rant about consulting buzzwords that hide real problems.
Real-Time Pipelines: When You Actually Need Them vs. When You Think You Do
Built a Kafka-based streaming platform that processes 2M events/sec. Cost: $40K/month. Actual requirement: data updated every 15 minutes. Sometimes batch processing is fine.