Discussing in terms of creating an AI Product, it has lot more to than Machine learning. Working on developing a ML model for a high tech company, I have learnt a lot apart from only Machine Learning which I am going to share here.
It is valid to state that by achieving 95% accuracy on a model using production data (which has a lot noise) is a great accomplishment. But, if you want to make it a sellable product and add value to increase your company’s revenue, you have to satisfy the customer segmentation predominantly. In order to get into the idealised state as a product, the solution has go through a lot of experts analysis to judge its current potential and to remodel it into valuable product forefront.
In current world, most the AI ML pipelines being automated and cloud deployed, it’s very important to deliver the best solution as per the client’s requirements. Sometimes it wouldn’t be a simple solution that might meet all the requirements, you might have to work on building entirely different solution to just satisfy your client needs. At the end, even if you have achieved 99% model accuracy, if company doesn’t see value in it or if the client rejected the deal, don’t get disheartened. At least you came know about the real product deals happening around.
From a Product Manager:
Working for a real life Product Manager with 20 years of industry experience, I have been learning a lot in the process of building a production level deployment ML models. He always asks us work and see through the client perspective with whatever we are progressing on. Like.
- Updating to use the ML model, Would I add value to current workflow ?
- By automating this step with Machine Learning, how far can we decrease manual effort ?
- How much revenue can I add to company if it can be sold as a product ?
- What could be its value when compared with the other competitors ?
- What is the feasibility and adoptability of the solution as a Product ?
etc… the list goes on.
Ultimately, we might meet few expectations and we might not. But, there is always a huge learning curve we progress along in delivering our best to make it scalable and sellable Product.
Keep exploring AI…..
The above content is completely based on my experience working from past 1.5 years. My work and perspectives might differ from company to company. Don’t take any facts and assumptions for granted.