The job of an AI/ML engineer seems like magic to those who are outside the field, as some of them perceive it to be that of a magicians, creating intelligent machines which can think and answer.
For starters, these folks may have FOMO (Fear of missing or missed out) as the technology world keeps updating with new newer models and frameworks. There was Anthropic’s Claude Sonnet which was a frontier model a while ago, now it is Mythos which is meant to be so powerful it can find zero day vulnerabilities within minimal time, which puts the entire IT infrastructure at risk of banks, financial institutions, governments and also individuals.
Once she/he overcomes this, real work begins, they start to look at deployed model performance metrics, meaning if the inferences are correct, any prompt injection attacks (in case LLMs) via a dashboard.
If this goes well, it is time to celebrate to at have it all food court, where everything is free and infinite expect for the appetite of the person. Now time to join the daily stand-up and wait for their turn while sipping coffee, give cryptic updates which only you can understand and ensure that the scrum master for sure can’t follow, if any questions do come up, give an answer which is even more tangential.
So far it has been stressful day, head out to the gaming room or start with instagram or linkedin feed for like an hour which feels like a minute, then suddenly remember an upcoming deadline. Login to workstation which is typically via ssh which has thousands of GPUs, hundreds of CPUs and upwards of 128GB RAM.
Now time to run a new training iteration and head out for lunch as the training and validation with test data will take few hours. After lunch a power nap to be woken by annoying MSTeams ping ! Grudgingly answer or ignore the ping based on who the question is from.
Now time to look at results , does not look any better than previous run, so write an email to Data team for even more data.
Well that has been a very tiring day, head out after another cryptic daily update email.
Automotive embedded software | Development engineering for software (personal views) 