Very insightful interview! Great work Walker.

Great discussion! Tulip for everyone to give frontline operators the tools they deserve

I hope Tulip is going to be at ProveIt! 🤞

WOW 👏 Tulip good job done !! Please Dont be expressive!! 1) What values in terms of *UNDERSTANDING* the data which was extracted from loads of batch record that the model ran through reading tons of PDF files of batch record? Can you please give OUT a few examples of such outcome you all have witnessed so far ? How many different models do they have or made so far ? No1. Reading the PDF/(paper based? ) or (even digital BRs? in which case do they need a specific format ?) Batch record and bringing out some value ( already asked above if you can describe all / some of those value outcome) No2. Model trained on Technical Specs - have they been generic or for a specific Vertical relating to the plant in which they trained these models. ( Re usability OR START from scratch for every installation) No3. Fault analysis based on (all alarms and events ? from historian ? ) and rectification actions taken to fix those faults. How does the operator describe the fault which model understands either ways even if his description is a bit Skewed by the operator. Requires a lot entry of information by operators , just like giving fault reasons as descriptions by the operators and cleaning that sensibly is another task. No4. *Walkers request* to know the scheduling of various SKUs BUT THAT is not a model , that is information already stored by the production planner in the system !! Its a simple query and nothing to do with LLM !! No5 *Composable MES* app !! INTERESTING need to really dive deep into it , see and learn !! but then which all Industry vertical has their model been trained on to throw out recommendations for ? Surely they would not have covered all possible Industry vertical and processes ( Batch, flow, continuous and custom, etc and then kind of Industry, steel , cement , f&b, pharma, petrochemical .. blah ..blah ) !! At the moment which all have their model been trained on to give good recommendations on!! Calling responses being *Hallucination free* ....NO !!.... may not be the right term; *Verifiable* YES !! but just because you provide reference to its responses, ( point to the teacher 🤣 BLAME IT ON RIO !!) The student STILL must be smart enough to know what is wrong !!!.. Anyways surely a smart move eliminating the cause of hallucinations slowly as the time progresses ( SMART THINKING) 😁

I didn't understand that part about machine learning and scheduling.