Avinash Kumar

Sr. Data Scientist & MLOps Engineer. Passionate about ML engineering & research. Fitness enthusiast seeking connections with like-minded folks.

[Avinash Kumar @iavinas]

151 days straight at the gym. No rest days. Here's what the data actually shows 🧵
@we_are_cult @bevel_health

[Avinash Kumar @iavinas]

@CommandCodeAI awesome 😍

[Avinash Kumar @iavinas]

@ctnzr @deepseek_ai @Xiaomi @OpenAI @MistralAI @aipoolside @NVIDIAAI I have recently switched from Opus 4.7 in Claude Code to DeepSeek v4-pro and Kimi k2.6. Both work really well, and I don't feel any difference compared to Opus. It just works.

[Pavel Durov @durov]

Even if you encrypt your WhatsApp backups with a strong password, your messages still end up in unencrypted cloud backups — because 90%+ of the people you message haven’t done the same. Add the fact that WhatsApp stores and discloses who you chat with, and the picture is dire.

[Avinash Kumar @iavinas]

People ask how I stayed consistent for 151 days.
Honestly? I always wanted to go. The gym became my therapy.
When your mental health is right, motivation is not the problem anymore. You just show up.
Cut the sugar. Cut the caffeine. Fix the mind. The body follows. 🔥

[Avinash Kumar @iavinas]

151 days straight at the gym. No rest days. Here's what the data actually shows 🧵
@we_are_cult @bevel_health

[Avinash Kumar @iavinas]

What that does to your engine:
VO2 Max: ~40 → 57 (Excellent)
BMR: 1410 → 1579 kcal/day
Walking heart rate trending lower for 22 straight weeks (now 84bpm)
Resting HR: 56.6 bpm
Your body gets more efficient at everything.

[Pratyush Kumar @pratykumar]

📢 Open-sourcing the Sarvam 30B and 105B models! Trained from scratch with all data, model research and inference optimisation done in-house, these models punch above their weight in most global benchmarks plus excel in Indian languages.

Get the weights at Hugging Face and AIKosh. Thanks to the good folks at SGLang for day 0 support, vLLM support coming soon. Links, benchmark scores, examples, and more in our blog - https://t.co/DcCG3zlN8p

[Chaithanya Kumar @ChaithanyaK42]

Over the last 1 month I have been really consistent with showing up for myself almost every single day & my current work out structure looks something like this

1. Currently prepping for a marathon that's happening on 22nd August ( will be my first ever) & planning to it under 4 to 4.5 hrs

2. Balancing running with strength training going with the saying, lift more than a lifter run more than a runner

Mon to Saturday I show up for myself everyday & I plan Sunday as usually a recovery day with still a min of 8 to 10k steps every day no matter what

I do strength training on tue , thur & sat & run on mon , wed & friday

Just one of month of consistently showing up has improved my overall health - been focusing on my hrv, rhr , biological age , as a researcher myself i genuinely care abt every goddamn metric & wanna health maxxxx & I am gonna do that

Current Priorities in life

Max out claude code & Codex limits, push my humanly limits on towards endurance training

DO HARD THINGS 😁

You can follow my journey here on strava -> Check out Chaithanya Kumar A on Strava
https://t.co/c02ANLi4Dx

[Avinash Kumar @iavinas]

@thepushkarp 🤣🤣

[Avinash Kumar @iavinas]

@julien_c 🤗

[Avinash Kumar @iavinas]

From your besties to your Aunties, it's so much funny 😆😭
Who at @calm decided Aunties needed a meditation subscription?
I respect it.

[Avinash Kumar @iavinas]

Learning a new topic with the help of LLM is so much fun. You can discuss and debate with it and finally understand it better.

[vLLM @vllm_project]

Great writeup from @AI21Labs on scaling vLLM for high-throughput, bursty workloads.

TL;DR: systematic config tuning + queue-based autoscaling = 2x throughput from the same GPUs. 🚀

Useful for anyone running vLLM in production with variable traffic patterns.

Thanks to the @AI21Labs team for publishing the full engineering writeup. 💙

🔗 https://t.co/FYBPRHGzHg

[Avinash Kumar @iavinas]

@IamKyros69 @brave

[Avinash Kumar @iavinas]

@AravSrinivas @comet Make it agentic over brave? I miss all the privacy features like fingerprint blocking and others.

[clem 🤗 @ClementDelangue]

We need better agent evaluations! Glad to have collaborated with @Meta Super Intelligence Lab to release Gaia2 and ARE!

GPT5 (high) from @OpenAI is leading on execution, search, ambiguity, adaptability and noise.

Kimi-K2 from @Kimi_Moonshot is leading open weight.

Full blogpost: https://t.co/KFSCdzdfwH

[Avinash Kumar @iavinas]

@ChaithanyaK42 Agree 💯

[Logan Kilpatrick @OfficialLoganK]

AI models are becoming agents out of the box