Skip to content
View Yug-Oswal's full-sized avatar
👨‍🎓
Currently (still) working towards the best ‘me’ I could think of.
👨‍🎓
Currently (still) working towards the best ‘me’ I could think of.

Block or report Yug-Oswal

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
Yug-Oswal/README.md

Hey! Nice to see you.

Welcome to my page!
I'm Yug, a 21 year old AI/ML developer/researcher and a 4th year UG CS student. (Also a self-proclaimed writer :))

Some of my research repos are currently private since they have some results our research manuscripts haven't yet updated. If you'd like to take a look please let me know.

Things I work with

PyTorch TensorFlow AWS Google Cloud Express.js FastAPI Flask Flutter NodeJS Firebase MongoDB MySQL Postgres Redis Keras Matplotlib NumPy Pandas scikit-learn Scipy GitHub JavaScript Python Docker

Where to find me

Website Github Twitter LinkedIn Gmail

A little more about me...

const yug = {
    pronouns: "He" | "Him",
    code: ["Tensorflow", "Numpy | Pandas | etc", "Javascript", "Python"],
    askMeAbout: ["AI/ML - best :)", "Research", "Data Science", "Web Backend", "App Dev"],
    technologies: {
        backEnd: {
            js: ["Node", "Express"],
            python: ["Flask", "FastApi"],
        },
        mobileApp: {
            crossPlatform: ["Flutter"],
        },
        devOps: ["Docker🐳", "CI/CD", "GPU Workloads for ML"],
        databases: ["PostgreSQL", "MongoDB", "Oracle SQL", "redis"],
    },
    ai/Ml: {
        research: {
            currentAreas: ["Novel Activations", "Loss Functions", "Neural Network Layers", "Theory of Deep Learning"],
            publishedPapers: 2
        },
        aboutMe: {
            projects: ["Rekindle (refurbishing)", "Duality of Weights", "Custom RNNs", "A few private repos (public soon)"],
            interesting: ["Trained a VGG-19 from scratch on ImageNet (1.5 million images) - 2 ArXiv Papers",
                          "Currently working and training on 2 NVIDIA RTX A5000s,
                           getting some A/L40s and trying to set up a distributed system for training really deep models 😁",
                          "Implementing my own RNN/GRU/LSTM currently, with novel cell state passing 😎",
                          "Interned for a US MIT Startup and BDL (Ministry of Defence, India)"],
            goals: ["Hmm... MS/CS-AI sounds interesting"]
        }
    },
    currentFocus: "Growing, building, and growing again",
    funFact: "I'm currently trying to beat ReLU"
};

I love connecting with different people so if you want to say hi, I'll be happy to meet you more! 😊

Github Stats ⚡

Github stats Top Langs

Pinned Loading

  1. conditioning-reasoning conditioning-reasoning Public

    Attempt to condition biased reasoning in LRMs using unbiased outputs of the same reasoning chain in the same model.

    Jupyter Notebook

  2. complex-behaviours complex-behaviours Public

    Attempting to induce complex behaviours in LLMs (eg. hidden bias propagation) via phase-dependent multi-steering. If it works, will use it to develop novel complex behaviours and stress-test and en…

    Jupyter Notebook

  3. cknob-interpretability-vone cknob-interpretability-vone Public

    This is an interpretability extension of a previous research project. In that, I basically amplify the the probability of a model to act P(Safe) while reducing P(Unsafe).

    Jupyter Notebook

  4. cr-actpatch cr-actpatch Public

    Activation patching to test causality of a new safety improved model tuned via my newly proposed algorithm

    Jupyter Notebook

  5. Motorq-VIT-2024-DS-Assignment-Yug-D-Oswal Motorq-VIT-2024-DS-Assignment-Yug-D-Oswal Public

    Repository for the MotorQ Data Science Assignment

    Jupyter Notebook

  6. reward-hacking reward-hacking Public

    Validating the hypothesis of inducing reward hacking in LLMs via steering

    Jupyter Notebook