As interest in agentic AI grows, many teams jump too quickly from "LLM-powered feature" to "autonomous agent." The result is often unnecessary complexity, weaker control, and systems that are harder to test and maintain.
In practice, many successful AI applications begin with simpler patterns: a single LLM call, a routing step, a prompt chain, or a bounded workflow with clearly defined tool use. Agentic behavior can be powerful, but it should be introduced deliberately, where it adds clear value.
This talk presents a practical, software-engineering-focused approach to building production AI applications by starting with workflows before agents. We will look at how to choose the right level of autonomy for a problem, how to separate deterministic application logic from model-driven behavior, and how to apply orchestration patterns such as routing, chaining, and evaluation loops within a broader architecture.
We will also discuss how these patterns fit with structured outputs, guardrails, tool boundaries, and escalation paths, so that AI capabilities remain useful without becoming chaotic. The emphasis throughout is on architecture and implementation, not hype: how to build systems that are easier to reason about, safer to operate, and more maintainable over time.
Attendees will leave with a practical mental model for deciding when a workflow is enough, when an agent is justified, and how to build either one in a way that supports real production needs.

Target audience
- Software developers, architects, AI/ML engineers, and technical leads interested in deploying AI beyond cloud GPU environments.
- The talk is especially relevant to developers building applications for laptops, mobile devices, embedded systems, IoT devices, or other environments where compute, memory, power, or network connectivity are limited.
- No deep machine-learning expertise is required, although familiarity with basic AI/ML concepts will be helpful.

Suggested Tags
Edge AI, AI/ML, Small Language Models, SLMs, LLMs, Quantization, Model Optimization, Embedded AI, On-Device AI, CPU Inference, AI Architecture, Machine Learning, Generative AI

Short tags list:
Edge AI, AI/ML, Small Language Models, Quantization, AI Architecture

Compact Version
AI is increasingly moving from the cloud to laptops, mobile devices, embedded systems, and other edge environments where memory is limited, GPUs may not be available, and connectivity cannot always be assumed.
In this talk, we will explore how to design AI applications specifically for these constraints. Rather than simply shrinking a cloud-based model, we will examine how to choose between traditional machine learning, compact transformer models, small language models, and cloud services; reduce resource requirements through quantization and low-bit inference; and build architectures that combine multiple models, caching, routing, and selective cloud fallback.
Through practical examples and benchmarks, we will see how model choice and system design affect latency, memory usage, and quality.
The goal is simple: build the smallest, fastest AI system that can reliably solve the problem.