Investing in “AI” or, more realistically, Machine Learning is the fashionable thing to do these days (in addition to investing in SPAC’s, NFT’s, etc.). But does your fledgling start-up really need a Machine Learning solution? Vinay Muttineni MBA2021 helped build scalable machine-learning pipelines and products at eBay and Microsoft. He highlights some of the things you should consider before going all-in on ML.
I have worked on a few Machine Learning products and, more importantly, have failed spectacularly in the process a few times. Before embarking on a ML journey I would ask these three questions:
1. Do you need ML/AI?
What is the objective in incorporating Machine Learning into the workflow? Is it just to attract VC money (Having “AI” in a pitch deck can help attract $$s) or is there more to it?
There are different flavours of Machine Learning projects. Common scenarios include:
- Increasing the perceptive power of your products: Using ML on sound, images, and videos to make your products more intelligent. For example, adding a conversational AI bot to your website.
- Adding predictive power to your product: Using ML to predict future outcomes. For example, making predictions on users’ spend patterns if you are a credit card company.
- Auto pattern detection: Using unsupervised techniques, you can segment your users better and provide more relevant information to them.
Whilst ML can be a powerful tool in your arsenal, it might not be the most effective one. For instance, a start-up I advised wanted to use ML/computer vision in retail to detect, identify, and count the products by taking pictures of the boxes they come in. Now, if you were an Amazon or a Microsoft and were inventing the future of shopping, this would be right up your alley. However, if inventing new ways of shopping is not your primary motive, you might want to look at a more simple and effective solution like using the barcode.
2. Do you have a clearly defined problem statement?
Once you have determined that using ML is going to be both an effective and efficient solution for your problem, you should then go about defining an exact problem statement. Speaking from experience, ML initiatives can quickly devolve into exploratory or research projects if a clear problem statement is not defined upfront and if the project is not well planned out with clear deadlines for each stage.
3. Do you have the resources? Data, people, etc.
Once you have a clear problem statement, it would be worthwhile thinking about the following requirements:
- Data: You need historical data to build ML models. Do you have enough relevant data to build a model? If you do not have historical data, do you have the means to generate a dataset via simulations and/or with labels generated by mechanical Turks?
- Employees: >70% of the effort required in building an ML model is taken up in the data collection and transformation stage. You would need a data engineer to own this piece and an ML person to strategize about the algorithms that would be most effective for your scenario.
- Tech stack: Most of the public cloud providers have readymade tools that would allow you to build out a simple Machine Learning pipeline very easily. Most even have visual and auto ML tools that allow non-ML engineers to build models. Use these to build your POC’s if you do not have the resources to hire ML engineers just yet.
This is part 1 of a three-part blog series. In the next editions, I will walk through the seven stages of an ML project through an example. In the final edition, I will leave you with tips around running an ML project.
If you liked this, you might like Five truths about algorithms you should know.
About the author: Vinay Muttineni has over 6 years of experience in Silicon Valley. There he helped build scalable machine-learning pipelines and products at eBay and Microsoft. He has a Masters in Computer Science from the University of Illinois at Urbana-Champaign. Now he is completing an MBA at London Business School. He has also worked in product roles at nPlan and Ori Biotech.
Previously, he volunteered at various high schools in the Bay Area teaching Computer Science. Currently he is mentoring students from non-target CS colleges through the Mentors in Tech program.