Skip to content
View sortira's full-sized avatar
🤔
poking into model brains *poke* *poke*
🤔
poking into model brains *poke* *poke*
46 followers · 59 following

Achievements

Achievement: QuickdrawAchievement: Pull SharkAchievement: YOLO

Achievements

Achievement: QuickdrawAchievement: Pull SharkAchievement: YOLO

Highlights

  • Pro

Block or report sortira

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
sortira/README.md

aritro shome

currently: @ ai4bharat, poking into multimodal multilingual open frontier models like gemma and analysing their performance on reasoning tasks like solving math problems across modalities and languages from an interpretability and alignment perspective.

personal projects: studying reinforcement learning and model red teaming, implementing projects of the same.

  • topics of interest: interpretablity, alignment and safety research, reinforcement learning, reasoning models, model red teaming.
  • more deets about me on my homepage
  • i also have a technical blog and a non-technical blog.
  • currently a research intern @ ai4bharat, iitmadras (mar 2026 - present)
  • ex-intern @ sarvam on their dubbing team (aug 2025 - dec 2025)

tech stack: python, huggingface, pytorch, numpy, pandas, wandb.

how to contact me:

(decreasing order of response times)

webrings i am a part of

threadlocked webring badge

Pinned Loading

  1. seven-deadly-sins-of-gemma seven-deadly-sins-of-gemma Public

    exploring the emotion vectors of seven deadly sins in representation space of gemma-2-2b

    Python 4 1

  2. forgeformer forgeformer Public

    pen is mightier than the GPU

    HTML

  3. alpona-gen alpona-gen Public

    a mathematically alpona generation tool to create datasets for training models on them

    Python 2

  4. thomas-raw thomas-raw Public

    a travel planner built using Langchain

    Python

  5. transformer-decoder-only-memorisation transformer-decoder-only-memorisation Public

    This is a POC of how transformers can be used to deliver payload by intentionally overfitting them to memorise the payloads and transmit as weights

    Python 1

  6. TheSopranos TheSopranos Public

    multimodal AI application with TTS, STT, depth and object detection for helping the physically challenged

    Python 1 1