As I reflect on my experience developing an AI product in 2022, I am struck by the numerous lessons I learned along the way. From technical challenges to business strategy, there were many obstacles and opportunities that required me to think critically and adapt my approach. In this blog post, I want to share some of the key lessons I learned from my experience in the hope that they might be of value to others working in the field of AI. Whether you are a seasoned developer or just starting out, these reflections and insights may provide some food for thought and inspiration as you embark on your own projects.
The importance of talking to user
One of the most important lessons I learned while developing an AI product was the importance of talking to the users early on in the process. As a developer, it can be tempting to focus on creating a product that is feature-rich and visually appealing, but this can sometimes be at the expense of the user’s needs and preferences. In my experience, I was building a dashboard for an internal organization that was full of colors and smart features for viewing data. However, when I presented it to the users, I quickly learned that they actually preferred to download the data as an Excel file and create their own dashboards.
This was a valuable lesson for me, as it reminded me that the role of the tool is to serve the user, not the other way around. By understanding the nature of the user and the specific needs and preferences they have, I was able to focus my development efforts on creating a product that was more tailored to their needs. In this case, the users were analysts who were very skilled with Excel, but needed help with preprocessing data at scale. By focusing on how to make this process faster and more effective for them, I was able to create a product that truly served their needs and made their work easier.
Data Science should be part of Engineering Team
Another lesson I learned while developing an AI product was the importance of considering the development process as a whole. In 2022, I was involved in multiple freelance projects as a data scientist. Most of the time, my progress was slowed down because the engineering team had already planned out a roadmap for new features. So my part needed to wait until other tasks were completed.
From this experience, I learned that data science should be integrated into the engineering process if you want to build smart features that truly make a difference. Project managers need to understand that data science often provides abstract ideas, but the implementation of those ideas is mostly carried out by the engineering team. This means that data scientists need to be more than just specialists in their field – they also need to be generalists who are equipped with the skills and knowledge to build features for websites and apps. As the world becomes more reliant on AI, it is essential that project managers have a thorough understanding of the development process for software that incorporates AI.
Supercharge value creation by not building everything yourself
Another lesson I learned while developing an AI product was the importance of utilizing services rather than building everything from scratch. While this may not be true for all products, I have found it to be particularly valuable in the software, research, and healthcare industries. For example, I once built an AI-powered SaaS platform to help digital marketers write better social media content. Initially, I was planning to build everything myself, from the frontend to the backend to the AI itself. However, as the user base grew, I was quickly bombarded with lots of bugs and other issues that took up a lot of my time and resources.
This made me rethink my roadmap and priorities. I realized that the main selling point of my product was the AI itself, and everything else outside of that focus should be handled by services as much as possible. This meant moving from building my own authentication to using a service like Supabase, from struggling with CSS to using Material UI components, from using a bare DigitalOcean droplet to using App Engine, and from building my own service to sync the database and email provider to using Pipedream. By making these changes, I was able to supercharge my development and create more value for the user faster.
Overall, my experiences developing an AI product have definitely helped me become a better engineer and project manager. I am grateful for the opportunity to learn from real-world challenges and successes, and I hope that these experiences will continue to serve me well as I work to bring value to my company. While every AI project is unique and presents its own set of challenges, I believe that the lessons I have learned will be applicable to many other situations and can help others avoid common pitfalls and achieve success in their own projects.