Skip to content
View himanshushukla12's full-sized avatar
πŸ’­
I may be slow to respond.
πŸ’­
I may be slow to respond.
  • Siemens digital industry
  • Bangalore
  • 14:49 (UTC +05:30)

Block or report himanshushukla12

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.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
himanshushukla12/README.md

Hi My name is Himanshu Shukla

Open-Source Contributions

My Professional Experience

ML Engineer at Siemens Digital Industry Software (July 2024 - Present)

πŸš€ Advance techniques for LLMs: Working on improving efficiency of Fine-tuning and RAG using some opensource tools.

  • Developed a contextual retrieval mechanism using LangChain, combining overlapping chunks, cosine similarity for top-K filtering (k=100), and BM25 retriever for optimized RAG performance. This is a concept implemented from Anthropic blog Idea.

AI Intern at Siemens Technology (Jan 2024 - July 2024)

πŸš€ Automation Wizard: As an AI Intern at Siemens Technology, I wield the power of algorithms to automate simulations, transforming complex processes into elegant solutions with the magic touch of prompts.

πŸ’‘ Innovative Solutions Architect: Crafting cutting-edge automation techniques, I collaborate with a team of visionaries to revolutionize simulation workflows, paving the way for efficiency and excellence in engineering.

πŸ”§ Problem Solver Extraordinaire: Armed with deep learning expertise, I tackle intricate challenges head-on, unraveling complexities and streamlining simulations with precision and finesse.

πŸ’» Advanced Techniques Specialist: Implemented multi-GPU training utilizing fault-tolerant mechanisms with TorchRun in PyTorch and worked on Graph Neural Networks (GNNs) to enhance simulation accuracy and efficiency.

Certifications

πŸŽ“ Deep Learning Maestro: Proud graduate of the Deep Learning Specialization by DeepLearning.AI on Coursera, equipped with the knowledge and skills to conquer the realm of artificial intelligence and reshape the future.

Deep Learning enthusiast (btw I also did android development)

I'm currently pursuing a Master of Engineering from BITS Pilani. Passionate about technology and always eager to learn new things!

Open-Source Contributions

  • Fixed RuntimeError: probability tensor contains either inf, nan or element < 0 link
  • Added Python script to run it completely locally using HF models in RAFT link
  • Fixed code error while doing evaluation in Contextual AI repository link

Skills

C C++ Git Java Python MySQL PyTorch Docker TensorFlow Machine Learning

Socials

Badges

My GitHub Stats

himanshushukla12's GitHub stats

πŸ“˜ What I'm Focused On

Currently, I'm honing my problem-solving skills by working through challenges on GeeksforGeeks, with a particular focus on Data Structures and Algorithms (DSA).

πŸ“ˆ GitHub Stats

Himanshu's Contribution Chart

πŸ“¬ Get in Touch

Pinned Loading

  1. ShishirPatil/gorilla ShishirPatil/gorilla Public

    Gorilla: Training and Evaluating LLMs for Function Calls (Tool Calls)

    Python 11.5k 998

  2. ContextualAI/gritlm ContextualAI/gritlm Public

    Generative Representational Instruction Tuning

    Jupyter Notebook 565 40

  3. text-to-image-stableDiffusioin text-to-image-stableDiffusioin Public

    end-to-end pipeline for generating a 4096 x 4096 image from a text prompt describing a person and their background. And I ensure, this will cover all the details of my work.

    Jupyter Notebook 1 1

  4. huggingface-llama-recipes huggingface-llama-recipes Public

    Forked from huggingface/huggingface-llama-recipes

    Jupyter Notebook

  5. llama-recipes llama-recipes Public

    Forked from meta-llama/llama-recipes

    Scripts for fine-tuning Meta Llama with composable FSDP & PEFT methods to cover single/multi-node GPUs. Supports default & custom datasets for applications such as summarization and Q&A. Supporting…

    Jupyter Notebook 1