PinnedBala Keelapudi·Feb 8Understanding AgentCore Observability: Session IDs, Traces, and Spans in ActionBuilding AI agents is exciting, but understanding what they’re doing in production is critical. This post walks through Amazon Bedrock…
PinnedBala Keelapudi·Nov 23, 2025Agentic IDE Kiro talks to AgentCore to ideate, create, deploy Agents to productionThe Problem we are solving
PinnedBala Keelapudi·Jun 26, 2025The “Meta-Agent” Design Pattern: Your AI’s Personal Assistant BuilderImagine you’re the manager of a super cool tech company, and every day people come to you with different problems they need solved. One…
Bala Keelapudi·Jun 27From Agent Failures to Engineering Specs: Building the GEDD Power for KiroYour AI agent hallucinates, You know this because your domain expert spent hours annotating failures — naming them, ranking severity…
Bala Keelapudi·May 31Evaluation-Driven Development for Agentic Systems: A Stratified Sampling ApproachOn the statistical foundations of domain-grounded evaluation, the precision-recall tradeoff in test set design, and why human-in-the-loop…
Bala Keelapudi·May 31Benchmarking AI Models the Way That Actually Matters: With Your Domain Expert’s Golden DatasetWhen you’re choosing a model for your AI agent, you look at benchmarks: MMLU, HumanEval, GPQA. They tell you which model is “smarter.” But…
Bala Keelapudi·May 7Grounded Theory for AI Agent Error Analysis [Part 1] :How qualitative research methods turn agent failures into systematic improvements, A Walkthrough with the SF Tour Guide
Bala Keelapudi·May 6The Grounded Theory Perspective — Building a Theory of Failure patterns for AI AgentsA practitioner’s guide on Qualitative Analysis for eval engineers
Bala Keelapudi·Apr 13Building an AI Customer Support Agent with Amazon S3 Files and Amazon BedrockCustomer support is one of the highest-impact use cases for LLMs. But building a production support chatbot usually means stitching…
Bala Keelapudi·Apr 12Spec-Driven Evals for Product Managers: How to Measure What Your AI Actually DoesWhy “it feels good” is not a shipping decision — and what to do instead