Building Successful AI Apps: The Dos and Don’ts

As businesses and organizations scramble to find good use cases for AI, several crucial questions consistently emerge: do you even need AI-powered tools? How should you go about building or integrating them into your existing workflows? And how will you know if the effort was worth it?

Whether you’re an independent practitioner or part of a larger team trying to make sense of this emerging technology, you’ll find concrete and actionable insights in the lineup of articles we’ve selected this week. They each tackle the nuts and bolts of building AI apps and leveraging their potential for well-defined goals, while avoiding common pain points.

While these posts zoom in on specific topics and business problems, they all offer a pragmatic, accessible approach, making them useful for readers across a wide spectrum of backgrounds and experience levels. Let’s dive in.

• What Did I Learn from Building LLM Applications in 2024? — Part 2
  So you’ve built a prototype of an LLM-based app that looks promising. What’s next? Satwiki De offers a comprehensive roadmap based on the takeaways she’s accumulated over a year of experimentation—the main thrust being to “develop your AI-enabled application keeping the business objectives in mind.”
• Why Generative-AI Apps’ Quality Often Sucks and What to Do About it
  Having observed countless enterprise AI initiatives deliver underwhelming results (if not fail altogether) Dr. Marcel Müller focuses on the importance of robust assessment, and shows “how we can qualitatively and quantitatively evaluate generative AI applications in the context of concrete business processes.”
• Designing, Building & Deploying an AI Chat App from Scratch (Part 1)
  If you’re ready to roll up your sleeves (real or proverbial) and start implementing an AI chat app, Joris Baan’s patient guide is a great resource to help you stay on the right track. Part 1 outlines the microservices architecture and local development needs you’ll want to think about, while part 2 moves on to a detailed discussion of cloud deployment and scaling.

Photo by Krišjānis Kazaks on Unsplash

• Learn to Build Advanced AI Image Applications
  Moving beyond chatbots, Ida Silfverskiöld explores the possibilities that visual generative-AI tools open up for real-world businesses and professionals—in this case, an interior design app based entirely on open-source models and frameworks.
• How to Build an AI Agent for Data Analytics Without Writing SQL
  Why not harness the power of AI to streamline common data-analytics workflows that would typically require numerous SQL queries? Chengzhi Zhao aims to accomplish just that with the aid of an AI agent, and shows how you can build one yourself with LangChain and DuckDB.
• AI-Powered Information Extraction and Matchmaking
  For another promising use case where an AI app can help you and your team save time and become more efficient, don’t miss Umair Ali Khan’s patient tutorial on building AI-based match-making tools (in this case, helping job seekers find positions that fit their skills and interests).

Branching out into the world beyond AI apps, we’ve selected a few more recommended reads we thought you’d enjoy—from a beginner-friendly intro to LLMs to an in-depth analysis of data strategies.

• If your less tech-savvy colleagues could use a clear and accessible primer on what LLMs are and how they work, just send them Carolina Bento’s top-notch introduction.
• For a more advanced exploration of LLMs’ inner workings, head right over to Jaemin Han’s fascinating look at models’ shortcomings in generating and interpreting ASCII art—and the surprising security risks the latter might suggest.
• Quick and focused, Clara Chong’s guide to variable scoping explains why this particular aspect of your code can have far-reaching consequences for your data science workflows.
• What rules should we have in place to protect humans from unknowingly interacting with AI profiles on social platforms? James Barney unpacks the fallout from the recent controversy around Meta’s (now-suspended) experiment.
• The rise of code-generating chatbots leads Murtaza Ali to wonder if the age of human-written programming tutorials is coming to an end. (Spoiler alert / sigh of relief: not just yet!)
• If you’d like to make sense of the gap you observe between between predictions and real-world outcomes, Hennie de Harder’s new post walks us through the basics of prediction probabilities, calibration, and how to interpret these numbers in a practical context.
• In the mood for a deep dive? Jens Linden, PhD just shared the latest installment in his Demystify Data Strategy series, focusing this time on the common pitfalls of data and AI strategies—and how organizations can avoid them.
• To round out this week’s lineup, we invite you to learn about MicroPython and its potential relevance for data scientists; Sarah Lea’s concise guide will get you up to speed on all the essential details.

Thank you for supporting the work of our authors! As we mentioned above, we love publishing articles from new authors, so if you’ve recently written an interesting project walkthrough, tutorial, or theoretical reflection on any of our core topics, don’t hesitate to share it with us.

Until the next Variable,

TDS Team