Prahitha Movva

Inference Engineering for Agents: Spend Compute where it Helps

To optimize AI agents, developers should practice "inference engineering" by focusing on the cost per completed task rather than simple per-token pricing. The article recommends al…

Loop Engineering: New Idea or Old Automation?

This article explores the concept of "loop engineering," a practice where developers design systems that allow AI agents to run, iterate, and self-correct autonomously rather than …

Prefix Caching for Builders

Prefix caching can reduce AI agent costs by up to 80% and improve latency, but improper configuration—such as placing dynamic data like timestamps at the start of a prompt—can inad…

How Leading AI Teams Design Evals for Production Agents

As AI transitions from single models to autonomous "fleets" of agents, traditional testing proves insufficient because real-world user behavior is too unpredictable to be fully cov…

Reinforcement Learning 101

This article explains that reinforcement learning (RL) is the foundational training method driving modern AI, enabling models to transition from simple pattern matching to complex …

Designing Agentic Memory in 2026

The article argues that building effective agentic memory requires a sophisticated architecture that distinguishes between five distinct cognitive memory types rather than treating…

Agent Skills Work But The Research Shows Most Teams Are Building Them Wrong

Recent research indicates that while AI agent skills can significantly improve task performance, human-curated skills are far more effective than model-generated ones, which often …