Visitar URL original
ssd0247 (Shubham) · GitHub
Skip to content
View ssd0247's full-sized avatar
🎯
Focusing
🎯
Focusing
3 followers · 9 following
  • 16:56 (UTC +05:30)

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Block or report ssd0247

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.

Content in all repositories owned by your account will be closed.
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
ssd0247/README.md

ABOUT ME

Hey there! 👋, Shubham this side 👨, aka ssd0247, Namaste 🙏

👀 I’m interested in all things data and math! The messier to make sense of, the better it gets 🤓

🌱 I’m currently learning and experimenting with PLs (Go, Rust), tools & frameworks (PachyDerm, PyO3, DVC, MlFlow, ModelChimp etc.) and how it can help me build and run CPU-intensive ML algorithms.

My search for collaboration is heavily skewed towards projects that bridge the gap between the current state of AI systems in-the-wild versus the marvels that are conjured in controlled environments of research labs. Engineering software systems that can think, act, reason and plan on their own accord. Upgrading from narrow-AI to general-AI which I expect, as per my current knowledge, will require both the symbolic-AI and deep-learning based approaches.

Topics that I have a fascination for is

  • AI-alignment (what it even means to align AI agents. Do we have a clear picture of what our own alignments are ? Can we bucket ideologies and beliefs under one roof ?)
  • AI-Embodiment
  • LLM hallucinations and ways to control and minimize its occurrence.
  • AI safety : In the end, there are real humans at the other end of these systems we build and therefore, all gaurdrails need to be deliberately enforced.

I am dogmatic about the belief that every software built by us, teaches us something enroute. So feel free to reach out to discuss any idea, no matter how naive it seems. Naive-Bayes algorithm is that only, "naive", but it still solves many real world problems. 😉

LANGUAGE AND TOOLS

Tensorflow PyTorch JAX sklearn spacy

Python Go Rust javascript

Pinned Loading

  1. filtering-engines filtering-engines Public

    Sub-projects aimed at conceptualizing, designing, and building recommendation engines, for in whichever sector or domain the need for filtering arises. Also contains traditional logic based algorit…

    Python

  2. LDALab LDALab Public

    A collection of heuristics, algorithms, tricks & techniques to support the NLP workflow involving Latent Dirichlet Allocation (LDA) technique

    Python

  3. nlp-engines nlp-engines Public

    NLP agents like general-query chatbots, simple chatbots, Named-Entity-Recognizer, POS-Tagger, text-summarizers and plethora of other NLP-tasks solved using computation. Application-driven modular &…

    Python

  4. object-detection-algorithms object-detection-algorithms Public

    A repo serving as a memoir of all the object detection algorithms I have come across or used in one project or the other.

    Python

  5. static-site-generator static-site-generator Public

    A static site generator. Tailored specifically for building Data Science related portfolios. Built using Go. Not a framework itself, but inspired from the giant named HUGO.

    Go

  6. Ramblings on Deep Learning Ramblings on Deep Learning
    1 
    This is a collection of gists, meant as a pedagogical tool in my own understanding of how the SOTA neural networks use mathematical theories of optimization & linear algebra (on the research side) & computer science concepts like quantization, cache, data-sharding, distributed computing, HPC, etc. (on the engineering side) to build models of intelligence (that happen to have emergent capabilities) & related service's infrastructure for consumption by millions of users.
    2 
    
              
    
              
    
                3
                
    These are standalone scripts and greenfield projects & in no way, production-grade. Integrating these in existing infrastructure stack wouldn't be a pleasent experience.
    
              
    
              
    
                4
                
    
              
    
              
    
                5
                
    Sometimes, GOFAI (symbolic AI) is also included, with encoded domain expertise & provides a more direct inductive bias (than high-level one of neural networks) of generating rules that take inputs to outputs. Often at scale & with edge deployments, the simpler the solution, the better the chances of not banging your head against the wall at 3AM in the morning.
    
              
    
          
    
    
    
  
    

    7.