| Item | Details |
|---|---|
| Event Name | AI Agents, Prompt Engineering, and AIoT Projects on AWS |
| Date & Time | 14/03/2026 |
| Location | floor 26 Bitexco Tower |
| Role | Attendee (FCJ Cloud Intern) |
This session introduced the limitations of standalone LLMs and explained why AI agents are needed in modern systems. Instead of only generating text, AI agents can perform multi-step reasoning, use tools, access external services, and make context-aware decisions. The presentation also explained the workflow of Strands Agents, including built-in tools, tool calling, system prompts, and knowledge base support. A live build/demo helped illustrate how agents can be created more efficiently with Strands compared to manual orchestration. :contentReference[oaicite:6]{index=6}
This session focused on how better prompts lead to better AI results. Key ideas included the importance of clarity, role definition, instruction quality, output formatting, constraints, and examples. The speaker also highlighted common prompt engineering principles such as being specific, separating sections clearly, avoiding ambiguity, and allowing the model to admit uncertainty when needed. In addition, the presentation covered token economics and the impact of prompt quality on usage cost. Advanced techniques such as Chain-of-Thought, Tree-of-Thoughts, RAG, and role prompting were also introduced, together with a browser extension called Proptimizer for automated prompt optimization. :contentReference[oaicite:7]{index=7}
The AIoT session presented a practical Locker Management project built with hardware devices and AWS services. The goal was to automate club locker usage, reduce manual processes, and improve asset tracking. The system used devices such as Raspberry Pi, Arduino, reed switch sensors, RFID card readers, LCD displays, and a camera. On the AWS side, services such as AWS IoT Core, Lambda, DynamoDB, S3, Amplify, and Rekognition were used to connect devices, store data, detect locker events, and perform face recognition. An additional project, Plutus: Financial Budget App, was also introduced as another example of applied system design. :contentReference[oaicite:8]{index=8}
One of the most important messages from the event was that AI systems are moving beyond simple prompting. Agents represent a more advanced model where AI can reason through steps, call tools, retrieve information, and act with greater autonomy. This makes them more suitable for real-world workflows that require integration, decision-making, and task execution. :contentReference[oaicite:9]{index=9}
The prompt engineering session made it clear that prompt design is not just a writing skill, but a technical skill that directly affects output quality, consistency, and cost. A good prompt includes role, task, context, input, expected format, and constraints. Better prompts reduce wasted tokens and improve the reliability of AI-generated outputs. :contentReference[oaicite:10]{index=10}
The AIoT project showed how AWS services can be combined with edge hardware to create intelligent, automated systems. AWS IoT Core enabled secure device communication, while Lambda, DynamoDB, S3, and Rekognition handled event processing, storage, and identity recognition. This demonstrated how cloud services can support scalable and practical AIoT solutions. :contentReference[oaicite:11]{index=11}
The Strands Agent session provided useful ideas for building workflow-oriented AI features. In future project work, agents could be used for tasks such as log analysis, multi-step report generation, or automated internal support workflows where the system needs to combine reasoning and tool usage. :contentReference[oaicite:12]{index=12}
The prompt engineering session can be applied immediately to any project that uses LLMs. By designing prompts more clearly and systematically, it becomes easier to improve response quality, reduce ambiguity, and optimize token usage. This is especially useful for AI-assisted reporting, summarization, and automation features. :contentReference[oaicite:13]{index=13}
The AIoT project offered a practical example of how AWS services can be combined into a real system involving sensors, identity recognition, cloud messaging, and web monitoring. This helped strengthen understanding of how to connect hardware, backend logic, storage, and AI services into one complete architecture. :contentReference[oaicite:14]{index=14}
This event was valuable because it connected foundational AI concepts with practical implementation. The progression from AI agents, to prompt engineering, to applied AIoT architecture made the learning experience both broad and concrete. Instead of focusing only on theory, the sessions showed how AWS services and AI techniques can be used together in realistic systems. The event also helped reinforce the idea that modern cloud projects increasingly require a combination of AI capability, system design, and operational thinking. :contentReference[oaicite:15]{index=15} :contentReference[oaicite:16]{index=16} :contentReference[oaicite:17]{index=17}
Overall, this event expanded my understanding in three important areas: AI agents, prompt engineering, and practical AIoT system design. It showed that building effective AI systems is not only about using a model, but also about designing workflows, structuring prompts well, and integrating the right AWS services. The sessions provided both conceptual clarity and practical inspiration that can be applied to future cloud and AI projects.