Skip to content

Latest commit

 

History

History
53 lines (41 loc) · 2.54 KB

experience.md

File metadata and controls

53 lines (41 loc) · 2.54 KB
Raw

Education

  • Indian Institute of Technology Kharagpur:
    • Integrated MSc in Economics: 2020 - 2025
    • CGPA - 8.6
  • Rajendra Vidyalaya:
    • High School: 2020
    • ISC - 95%
    • ICSE - 95.8%

Technical Skills

  • Skills:
    • Data Analysis, Web Scraping
    • Natural Language Processing, Machine Learning, Deep Learning
  • Programming Languages:
    • Python, C++, SQL, JavaScript
  • Libraries & Tools:
    • pandas, numpy, Selenium, scikit-learn, transformers, PyTorch
    • Colab, GitHub,
    • Stata, Microsoft Excel

Experience

Prompt Engineering Intern

OK2 Health Apps, Remote: Jan 24 - Mar 24

  • Created robust functional calling code for getting structured responses from ChatGPT, using the GPT-4 API
  • Created concise yet informative prompts to generate step-wise hints for helping children learn chess through small puzzles and tactics
  • Extracted information on positional imbalance and best possible moves using python-chess library and Stockfish
  • Implemented the described functionalities in a single Colab notebook to output structured data from various input types, such as lesson text, PGN or FEN string
  • Designed multiple such notebooks for generating tutor prompts for different puzzles types on the MyChess app

Data Science Intern

Indian School of Business Hyderabad, Remote, Remote: Aug 23 - Sep 23

  • Implemented scrapers to extract 3,000+ monthly district-wise rainfall and GDP records of Madhya Pradesh
  • Curated dataset of past election winners to assess the possible effect of GDP and poor rainfall on election results
  • Extracted 4,000+ Facebook comments on candidates’ posts to perform Sentiment Analysis of the public responses

Data Science Intern

Indian Institute of Management Calcutta, Remote, Remote: Jun 23 - Jul 23

  • Extracted 1,500+ YouTube video comments and formulated a multi-class training dataset to perform text classification
  • Preprocessed the dataset by translating Hinglish comments to Hindi and resolved class imbalance by upsampling
  • Performed fine-tuning on a pre-trained multilingual DistilBERT model from HuggingFace, achieving a weighted F1-Score of 0.791 with more than 50% accuracy in four labels

Data Analyst Intern

Tata Steel, Remote: Jun 22

  • Analysed top customers, customer domains and products by visualizing purchase frequency over time
  • Devised potential product combinations for individuals and companies based on buying history
  • Identified 3rd of the month as most active with peak buying hours around 2 PM to 10 PM, with an upward trend in sales after January