Rajat Kateja

Notes from FAST 2026 I attended the File and Storage Technologies (FAST) conference in Santa Clara this week. Some observations and thoughts: "Art is never finished, only aband...

Thoughts (mostly non-technical) from attending the FAST conference last week:

afterhoursacademic.com/fast-2026/

Note to self: start the next presentation with "Hey everyone, I am super nervous to present..." funny + true + relatable

A brief introduction to LSM trees Log-structured merge (LSM) trees are widely used in key-value stores and as the storage layer of database systems. There is a rich tradeoff space in LSM tree...

New blog post describing the basics of LSM trees afterhoursacademic.com/lsm-trees-in....

A case for putting myself out there, aka, how to be okay appearing stupid I recently faced some scenarios where I was afraid to put myself (more specifically, my thoughts/ideas) out there because of the fear of appearing stupid (i....

New blog post about learning how to put my ideas out for critique

afterhoursacademic.com/stupid-okay-...

Parallelism strategies for neural network training Any large-enough neural network training job (e.g., training a LLM or an image or video generation model) employs parallelism strategies. In fact, such large...

New blog post about the different styles of parallelisms available/used for training neural networks

afterhoursacademic.com/parallelism-...

ETTR, aka, how good is the training goodput Effective training time ratio (ETTR) defines the ratio of time spent in training to the wall clock time. In intuitive terms, it is the goodput of the trainin...

New blog post about the effective training time ratio of ML model training jobs

afterhoursacademic.com/ettr/

Flounder Mode Kevin Kelly on a different way to do great work

Great read!

I have strayed from the well-established path/ladder many times in favor of following my interests. It doesn't always work out, but it *always* leads to interesting experiences.

Of course, there is value in sticking though with things as well.

It's a balancing act!

Highs and lows of 2025 [+] Changed jobs and landed at one I like [+] Learning new and interesting things at work (about carbon emissions and LLM training/inference) [+] Lea...

New blogpost: highs and lows of 2025

How to make LLM checkpoint writes fast Unless you have been living under a rock for the last couple of years or so, you already know what a large language models (LLM) is. For the rock-dwellers, h...

New blog post covering the most prevalent technique and some new proposals for fast LLM checkpoint writes.

afterhoursacademic.com/llm-checkpoi...

Friendships I did not realize I was being bullied. Largely because I did not understand the concept of online trolling or bullying. This was 2010 after all. I was living...

afterhoursacademic.com/friendships/

I really appreciate and resonate with the anti-AI-use sentiment (refusing to use AI assistants to write emails/blogs/code) that I have been seeing recently.

Maybe I am just being old and refusing to learn new tech, but much of the new tech really feels useless...

people will make fun of ancient civilization for consulting oracles and cutting up livers to tell the future and then type “what is the meaning of life?” into ChatGPT

A employee-director or employee-founder relationship is somewhat similar. Yet, the gut feel check is not as prevalent. Instead, the project or work typically carries more weight.

Or maybe it was just me who was doing this wrong...

7/

The gut feel criteria is pretty widely known/accepted when it come to choosing a PhD advisor.

"Choose the advisor, not the project" is what I was told when starting my PhD.

Projects change over time and a good advisor-advisee relationship is a crucial for success.

5/

A casual hang out also provides a way to do a gut-check. These folks are key stakeholders in your trajectory at the company and hold immense power over you. Why join their team if you don't get a good feel about them.

4/

If the VP can only talk about work or thinks that they are the smartest person on the table, that's probably what it takes to succeed there, no matter what they say about work-life balance or growth mindset.

This is the "watch, not hear" equivalent of "show, not tell".

3/

The senior people are ones that have succeeded in that environment/company/culture. Chances are that succeeding in that place requires adopting a similar lifestyle.

2/

The "watch, not hear" corollary of "show, don't tell" when choosing a job.

I recently realized an underrated way to determine if a job is a good fit or not: hang out with senior folks in the team over a coffee/meal (principal engineers, VP, or even C-level folks depending on the company size)

1/

Vibe coding and prompt engineering sound like good ideas but are not.

Edgar Dijkstra's 1978 paper "On the foolishness of natural language programming" makes a great argument!