Kaustubh Singhal

About Kaustubh Singhal

Product Data Scientist with a sharp focus on solving high-stakes business problems using…

Experience

1. Product Data Scientist

   Maybank

   Aug 2022 — Present · Bengaluru, IN

   Designed and deployed a RAG-based chatbot for user guide search using FAISS, embedding-based retrieval, reranking, and Llama3, integrated with LangChain for conversational follow-up.2. Led NLP modules in the Name Screening pipeline:a. Applied Agentic LLMs (WebBrowser, SmartScraperGraph) for live article scraping and summarization.b. Used OCR (EAST, PaddleOCR, Passportmrz) to extract multilingual entity data (e.g, Chinese names from passports).c. Developed Named Entity Recognition (NER) using BERT, CRF, and LLM prompting (Llama3/3.1), achieving up to 78% F1 on entity-level tagging.d. Built transformer-based Q&A models to extract risk insights from articles with Phi3.5 and Llama3.1.e. Engineered a combined prediction model (Random Forest, XGBoost, SVM) for entity classification with 84% F1.3. Forecasted macroeconomic trends using multivariate time series models (SARIMAX, VARX, GARCH), achieving MAPE < 0.05 on key indicators.4. Built credit risk and sales growth models using both ML (XGBoost, Random Forest) and time series (ARIMA, Prophet), integrating financial data and macro indicators.

Education

• Indian Institute of Technology, Mandi

  Master of Technology - MTech

  2019 — 2020

• College Of Technology, G.B.P.U.A&T, Pantnagar

  Bachelor of Technology - BTech

  2014 — 2018

• Indian Statistical Institute, Kolkata

  Master of Technology - MTech

  2020 — 2022

• DAV Centenary Public School - India

  12th (Higher Secondary Certificate Examination)

  2013 — 2014