Category: Agentic Design Patterns
Generative AI vs Agentic AI vs AI Agents
We use the same word — AI — for three very different jobs. One writes your draft. One runs your process. One drives your truck. When teams say let’s add AI, they often mean one of these but build the wrong abstraction. Generative AI is a production engine. It turns a prompt into content. Agentic…
Read MoreAgentic Pattern: Agent-Specific Context and Memory
In enterprise multi-agent systems (MAS), treating context and memory as a single monolithic conversation buffer quickly breaks down. The Agent-Specific Context and Memory pattern decouples and scopes memory hierarchies – ephemeral working scratchpads, session state, and persistent episodic/semantic stores – to individual, specialized autonomous agents. Rather than maintaining an ever-growing global blackboard or concatenated dialogue transcript, each…
Read MoreAgentic Pattern: Sensing with RAG
In classical autonomous agent design (originating from control theory and cognitive robotics), an agent operates via a continuous loop: Perception (Sensing) → Cognition (Reasoning & Planning) → Action (Tool Execution) → Environment Feedback. While traditional Retrieval-Augmented Generation (RAG) is deployed passively to answer human queries, the ‘Sensing with RAG’ pattern repurposes semantic retrieval as an active, real-time sensory organ for autonomous agents. Rather than overloading the…
Read MoreAgentic Pattern: Single Agent Baseline
The Single Agent Baseline (SAB) is the foundational architectural pattern for autonomous and semi-autonomous AI systems. It models a single cognitive loop where a large language model (LLM) acts as a centralized reasoning engine equipped with: In modern agentic engineering, the Single Agent Baseline is the mandatory control benchmark. Before justifying the distributed complexity, non-determinism, and token overhead…
Read MoreWhy System Prompts Aren’t Security Boundaries: Moving to Zero-Trust AI Agent Architectures
When autonomous agents move from text generation to mutating production state, executing dynamic code, modifying ledgers, or triggering payouts, traditional perimeter firewalls and system prompts fall short. System prompts are soft constraints easily bypassed by prompt injection or model drift. To run agents safely in production, we must enforce deterministic security controls outside the LLM…
Read MoreAI Stack: Products & Software
The AI Stack: Products & Software reference maps Well-Architected pillars to modern AWS, Azure, and open-source technologies used to build production AI agents with MCP, RAG, and LLM systems. It compares ingestion, transformation, retrieval, inference, evaluation, and observability layers to help teams choose managed vs open architectures. Pillar Capability AWS Azure Open Source Operational Excellence…
Read MoreAgentic Pattern – Reflection notebook code
code: https://github.com/krishnamanchikalapudi/examples.ipynb/blob/develop/LLM/Agentic%20Patterns/Parallelization.ipynb
Read More