Manish Rai [00:35:42] Yeah. And I can take the question directed at me. It's a hard question, but I would like to address the elephant in the room. Thomas, I don't know which part you agree or disagree with my outlook, because what I said was the first generation IDP solutions were not very flexible. And I think that's exactly what you're saying over here, that you found that some of the approaches were not flexible. They had a fixed set of OCR. They're using it as a black box of EAI models inside and you don't have visibility and then you don't have a quality guaranteed associated with that. And I couldn't agree with you more on that point that yes, those solutions were not flexible. And our approach here and thankfully a lot of companies in the market, you know, they were not API first companies, they were in RPA or some other business and they put out that solution and they lacked a lot of deep expertize. Why is the difference with some areas of our foundation, our engineering team comes from people from Google Green team, from the original founders and the Microsoft brand himself as a ML BSD dropout from M.I.T.. And and so we have solid engineers who understand all the challenges and how to overcome right to where you could have over fitted model or you could have a loss by overtraining it and things like that, whatever can go wrong. And we have taken steps to make sure that we maintain the quality and improve the quality over time continuously and do the active learning in the right way possible without relying too much on the outliers and having them kind of affect our underlying models and impact results.
Somachine 31 Download Full
2ff7e9595c
Comments