Freshworks product

When you are trying to make something general Model which should serve the Coffe shop and A bank both, for eg you got two tickets one from the coffee shop and another from the bank, you should give as much meta data or supporting data as possible for eg take a json, which have the business details like it is a coffe shop, which services they opted, telling the moment when the login issue started kind of change point, when the software push happened, Customer footfalls is monotonic increasing hourly etc which gives the LLM more context or in case of ML, giving the mean, median and variance of 1 hours, this will give the model info that this is anamoly getting 50 login failures in 1 hour.spark can do this for you.

On scaling part: Spark can take all the tickets and dividie it into multiple parts and give to the models in different servers.

pipeline: Data → Analytics → LLM → Natural Language.
Taking data from snowflake, Spark analysises the central tendencies like mean and variance etc to more insights to the model and then LLM generates something more readable for eg. Hey your Bangalore customers are facing a lot of login issues.

Processing times: 12:AM Night in every timezone if done daily, same can be done on weekly or monthly or when a user signup
use the Async and divide the data to pass the multiple

If thousands of job got trigerred at one point, the SQS (Amazon Simple Queue Service) can help to process the data properly rather than all the requests hitting the database at once.

When we have data like tickets from in progress to solved, then we will have to rebuild the data for that company if the tickets are not independent else we can trigger the change in the database and the cache.