Welcome to my GitHub profile! I am Sarthak Sharma, a dedicated software developer with a keen interest in Android development, machine learning, and contributing to open-source projects. Currently, I am pursuing my Master's in Computer Science and Engineering at IIIT Delhi. I am passionate about leveraging technology to innovate and solve real-world problems. My journey in the tech world has been shaped by hands-on experience in developing diverse projects that span from multimodal data analysis to Android application development and machine learning model optimization.

Feel free to reach out to me for collaboration, questions, or just to say hi!

• Name: Sarthak Sharma
• Location: [New Delhi, India]
• Email: [email protected]
• Education: MTech. CSE, Indraprastha Institute of Information Technology, Delhi (2023 - 2025)

• Conference Name: Interspeech 2024
• Authors: Sarthak Sharma*, Orchid Chetia Phukan*, Drishti Singh, Arun Balaji Buduru, Rajesh Sharma
• Repository: Arxiv Link

• Description:
  • Ensembles the power of Audio Visual Data to generate humorous notions for the machine to understand.
  • Utilizes Pre-Trained AST (Audio Spectogram Transformer) and VideoMAE models for embedding generation.
  • Develops a CNN-based fusion model using TensorFlow and PyTorch.
  • LanguageBind model used as an alternate approach but less accurate than VideoMAE and AST models.
  • Experimented with several CNN and LSTM models, deployed the most accurate one over a Tkinter application.
• Technologies: TensorFlow, PyTorch, Tkinter
• Repository: GitHub Link

• Description:
  • Leverages the power of data-driven insights to predict accurate delivery ETA (Estimated Time of Arrival) using DoorDash dataset present on Kaggle.
  • Utilized the attributes already present in the dataset and applied feature engineering to get novel attributes to train the model instead of feeding the initial data.
  • Experimented with various Machine Learning techniques such as XGBoost, Random Forest Regressor, LightGBM, and among them XGBoost tends to be better among these regressors.
  • Metrics used were Mean Absolute Error(MAE), Mean Squared Error(MSE)
• Technologies: PyTorch, sklearn
• Repository: GitHub Link

• Description:
  • Developed a personalized travel web application for tailored travel plans.
  • Augmented City and Hotel Datasets for better model comprehension.
  • Explored RAG, Langchain, and other LLMs for relevant travel suggestions.
• Technologies: Python, Information Retrieval, LLM, NodeJS
• Repository: GitHub Link

• Description:
  • Implemented Linux commands such as cd, mv, rm, cp, and ls in C++.
  • Added multithreading for recursive options, measured performance improvements.
• Technologies: C/C++
• Repository: GitHub Link

• Description:
  • Developed a Resume Parser to extract key entities from resumes.
  • Utilized BERT for embedding extraction and custom models for entity identification.
  • Deployed using Flask for functional information extraction.
• Technologies: Python, Information Extraction, BERT, Flask
• Repository: GitHub Link

• Description:
  • Fine-tuned GPT2 and T5 pre-trained models for text summarization.
  • Used Amazon Fine Food Reviews Dataset for context-based fine-tuning.
  • Achieved significant results with T5 Model for shorter reviews.
• Technologies: TensorFlow, PyTorch, Python, T5 Model, GPT2 Model
• Repository: GitHub Link

Thank you for visiting my profile! Let's build something amazing together.