A R&D team working for me at a tech company developed an advanced machine learning algorithm that could detect duplicate records in large datasets with high accuracy. However, the R&D team struggled to turn this into an actual data anomaly detection service that could be used by the develop team responsible for production.
There were a few reasons for this:
1. The algorithm required very large amounts of data to train the ML model and achieve high accuracy. But the R&D team did not have easy access to large, high-quality datasets that dev team would find useful.
2. The algorithm and ML models were complex, and the R&D team lacked expertise in building a simple, intuitive user interface that could hide this complexity and make the service easy to use.
3. Strict data privacy regulations meant that customer data had to be kept confidential and not used to train the ML models directly. But the R&D team could not figure out a way to do model training and inference without accessing customer data.
4. The R&D team focused only on the ML algorithm but did not have the full set of skills needed to build an end-to-end service. They lacked engineering skills around robust production quality API development, web/mobile interfaces, security, and DevOps.
In summary, while the R&D team was able to develop an innovative data record duplicate solution, they lacked some of the skills, expertise to turn this into a useful, service. With more collaboration across teams, they may have been able to overcome these hurdles.