Dyotak Kachare

Education

Vivekanand Education Society's Institute Of Technology (VESIT), Mumbai Dec. 2021 – May 2025

Bachelor of Engineering in Artificial Intelligence and Data Science with Honours in Blockchain CGPA: 8.68

Institutional Shareholder Services - STOXX July 2025 – Current

Junior Analyst

Developed a proof-of-concept RAG application on Vespa.ai, enabling document ingestion and high-recall retrieval.
Implemented Python scripts to benchmark & compare Vespa.ai and Elasticsearch performance, validating accuracy.
Replaced Google Translate API with an LLM-based translation, implemented glossary feature for translation consistency.

Mercari, Japan December 2024 – February 2025

Machine Learning Engineer Intern

Worked in the Trust and Safety team on the Item Moderation project.
Worked with a highly imbalanced dataset (10^-6 positive class ratio) to improve model performance.
Enhanced the production LightGBM model, increasing precision from 3% to 5%.
Fine-tuned Japanese BERT and DistilBERT models, further improving precision to 6.5%.

Cere Labs May 2024 – July 2024

Software Engineer Intern

Developed a parser for html documents that extracts text from html pages to be ingested for chatbot.
Developed a service that extracts html pages related to user query and uploads it to another internal service.
Improved performance of existing website by changing a polling request to a websocket connection.

AI YouTube Search | PyTorch, BERT, Streamlit, Pinecone

Created a RAG (Retrieval-Augmented Generation) system for a limited YouTube video dataset.
Embedded video transcripts using sentence BERT and stored embeddings in Pinecone vector database.
Developed a user interface with Streamlit to facilitate user interaction and video search.

Credit Card Fraud Detection | Python, Flask, React

Developed a machine learning model for detecting fraud in a highly imbalanced credit card transaction dataset.
Implemented SMOTE and undersampling techniques to improve model accuracy.
Created a backend server using Flask and a frontend interface using React.

Languages: Python, Java, C, HTML/CSS, JavaScript, SQL

Technologies/Frameworks: Scikit-learn, PyTorch, TensorFlow, Flask, FastAPI, React, Node, MongoDB, Docker

Tools: Git, GitHub, MS SQL Server

AI CoLegion May 2024 – May 2025

President Vivekanand Education Society's Institute Of Technology

IIIT SriCity National Health Hackathon

Runner-Up

ISTE Invictus

Winner

VJTI Refactor

Participation

Google Developer Student Club - Bit n Build

Participation