Realisation
I got a chance to be a part of Interview for a Research Internship role where i will work on JEPA based world models.
The interviewer asked me questions. The questions I expected yet not fully understood. It was a blood bath for me because its been sometime I looked into basic concepts that a research internship interview expects me to know and I do know most of them but nearly forgot as I went through them a long time ago and never looked again.
I also need to go a lot deeper into the bread and butter of the field of AI : Transformers and Attention. I was unable to tear the parts down, explain each and could tell what happens if one part is substituted with another new part.
I was unable to answer a question that haunts me till now.
He asked “Lets say you are training a model and the loss saturates, what is your process of debugging” I never thought about this. As I was always following along and never thought to look back or like stop and think.
Get Maruthi Enugula’s stories in your inbox
Join Medium for free to get updates from this writer.
He suggested me to do this thing “ Question every choice you make when you are researching or building something” that way you get to understand low level stuff.
I was able to answer stuff but mostly high level answers, I felt my answers didn’t contain as much technical depth as one expects for a Researcher level.
I need to up my game.
I am starting a series ( blog / vlog / tweets / whichever most comfortable with ) and I will work on my research skills.
Till next time.
Maruthi.
