From chatbot to agent: the production checklist
Define tools, permissions, approval steps, and failure modes. Add evals, logging, and safe rollback.
Tags: agents, guardrails, observability
Insights
Practical engineering for Agentic AI.
Short, actionable posts built from delivery experience.
RAG that doesn’t hallucinate: what actually works
Use citations, curated sources, access controls, and “answerable/unanswerable” classification.
Tags: retrieval, citations, grounding
Threat modeling GenAI systems in enterprises
Prompt injection, data exfiltration, insecure tool calls—mapped to mitigations and monitoring.
Tags: OWASP, injection, audit logs
Shipping ML responsibly: evaluation and drift basics
Track datasets, metrics, bias checks, and post-deploy drift alerts with retrain pipelines.
Tags: monitoring, evaluation, governance
Cost-aware cloud architectures for AI workloads
Right-size compute, cache smartly, autoscale, use spot where safe, and enforce budget guardrails.
Tags: FinOps, scalability