Atharva Ingle (AI)
@AtharvaIngle7
Followers
2K
Following
10K
Media
180
Statuses
2K
Data Scientist II @Wolters_Kluwer , @kaggle Competition Expert, @weights_biases Ambassador
Pune
Joined November 2020
Excited to share that I'll be speaking at the @weights_biases MLOPs conference (Fully Connected) on March 15th. There are a lot of exciting talks by legends like @jeremyphoward @l2k @HamelHusain @borisdayma @JayAlammar and many more. [cont.👇].
6
1
40
Hi everyone,. I’ve decided to move on from my current company after almost 3 years and I am open to new opportunities in AI/LLM engineering/research. I have designed and deployed production-grade RAG systems, fine-tuned and served open LLMs at scale. Key highlight from my.
6
7
40
Grok 4 refused to run a huge prompt 💀 (Claude Opus 4 handled it lol)
1
0
2
this is really cool - like how it builds a personalized evaluation roadmap based on your specific situation.
I made a choose-your-own-adventure for AI evaluation strategy. Your answers build a personalized roadmap based on task complexity, cost of failure, and your current evaluation. It also includes my favourite selection of tips from industry experts like @eugeneyan, @chipro and
0
1
5
I was sleeping on claude sonnet 4 and opus 4 and man, they are really good models. I actually read their advice and the responses they give, rest of the models responses i just don’t feel like reading and skim through. Also very solid in tool calling, writing, explaining things.
1
1
4
RT @AtharvaIngle7: Really cool project and experiments. I went through this paper and here are my notes and takeaways. remember when @kell….
0
1
0
Sources:.- Andrej's tweet.- Paper:
Love this project: nanoGPT -> recursive self-improvement benchmark. Good old nanoGPT keeps on giving and surprising :). - First I wrote it as a small little repo to teach people the basics of training GPTs. - Then it became a target and baseline for my port to direct C/CUDA.
0
0
2
Really cool project and experiments. I went through this paper and here are my notes and takeaways. remember when @kellerjordan0 modded @karpathy GPT-2 repo (nanoGPT) and developed it into a speedrun challenge ? people iterated on it and slashed the training time from 45 mins to
Recently, there has been a lot of talk of LLM agents automating ML research itself. If Llama 5 can create Llama 6, then surely the singularity is just around the corner. How can we get a pulse check on whether current LLMs are capable of driving this kind of total
1
1
10
this is huge, great win for OpenAI.
thrilled to be partnering with jony, imo the greatest designer in the world. excited to try to create a new generation of AI-powered computers.
0
0
1
loved reading this, so wholesome vibes.
✨ This week I visited my API provider @FAL in San Francisco. It's funny to go visit the ACTUAL physical place where you send tens of thousands of CURL requests to every day (well kinda cause they host stuff not in the office of course). FAL is responsible for hosting ALL the AI
0
0
3
Was reading some old @BearNotesApp blog posts just for fun and came across this snippet. reminder that building a good, performant and pleasant editor is really hard. Honestly, this app still has the best editor I’ve ever used anywhere.
0
0
12
Wow, the 'Python, the Documentary' trailer is out and it looks really good, can't wait to see the film!.
The trailer for "Python, the Documentary" by @CultRepo premieres today! 🔥🐍. From its origins in the Netherlands to powering AI at top global companies, this is Python's story. 🎥 The film premieres at @EuroPython in July!. Subscribe to @CultRepo's YouTube channel for updates!
0
0
2
New Lil’Log just dropped, will be a long weekend read :).
Giving your models more time to think before prediction, like via smart decoding, chain-of-thoughts reasoning, latent thoughts, etc, turns out to be quite effective for unblocking the next level of intelligence. New post is here :) . “Why we think”:
0
0
6
Qwen’s vision models are really solid and great candidates for fine-tuning. I've been using the 7B vision model internally in our company for document information extraction, which is tricky in our case: data spans from the 1980s to 2025, with wildly different formats and edge.
I have seen that Hugging Face really values ideas. Today @mervenoyann pointed out that while reading the Qwen 2.5 VL paper she noticed that we could perform object detection with the model. @SergioPaniego and I were already cooking something similar with a different model, so
0
0
4
was looking to buy a new vehicle for my family — had like 10 models to compare. posted a screenshot of the pricing table for all the models into gemini 2.5 pro deep research. it went through like 100+ websites and gave me a full solid easy to follow report. what would've taken.
2
0
6