Hello, I'm

Debanil Bhattacharjee

I am currently working at the Decision Science team at Trust & Safety Division of PhonePe,Focused on developing data-driven strategies to enhance user safety and Combat Financial Fraud, with competencies in fraud detection, risk management, and data solutions, Machine Learning,CNN,RNN and Fine Tuning Of Large Language Models.My other experience includes fraud and risk management in consumer payments(Peer to Peer) as an intern at PhonePe.

Get In Touch GitHub

About Me

I work at the Decision Science,Trust and Safety Team at PhonePe where we make data and machine learnng based decsions on millions of users everyday to protect them from finacial fraud and cyber crime.

With expertise in big data technologies, machine learning, and fraud prevention, I've contributed to reducing false positives by 15% and fraud incidents by 20% through data-driven solutions.

2+

Years Experience

15%

Reduction in False Positives

20%

Fraud Reduction

Education

Bachelor of Technology in Computer Science

National Institute of Technology, Agartala

2020 - 2024

CGPA: 8.62/10

Experience

Analyst – Trust & Safety Decision Sciences

PhonePe Private Limited

June 2024 – Present | Bengaluru, India

Built a fine-tuned LLM-based model to generate real-time interactive summaries of user risk profiles, enabling investigators to quickly assess fraud involvement
Created end-to-end labeling system for entire user base with good/bad tags based on behavior, affinity, and identity for proactive threat identification
Reduced false positive blocks by 15% through comprehensive analysis of rule-based system variables while maintaining fraud prevention efficacy
Integrated Device-Card Tokenization for VISA and Mastercard Networks, reducing card phishing incidents
Optimized daily batch ingestion/Spark job DAG runs, reducing failures and spillovers through task parallelization and schema restructuring

Intern Graduate Trainee

PhonePe Private Limited

Jan 2024 – June 2024 | Bengaluru, India

Analyzed and implemented checks based on multiple risk signals in P2P payments, achieving 20% week-on-week fraud reduction
Created data stores and dashboards for workflow tracking, anomaly detection, and daily data quality governance

Summer Intern

National Informatics Center

May 2023 – July 2023 | Agartala, India

Created REST API using Spring Boot to convert HL7 ORU data from medical lab testing machines to PDF format

Featured Projects

End-to-End AI Sentiment Analysis Service

August 2025

Designed and implemented a production-ready text sentiment classification service using transfer learning with HuggingFace Transformers, training on 50,000+ labeled movie reviews.

Fine-tuned DistilBERT achieving F1 score of 0.93
Built reproducible data pipelines with real-time API response
Containerized with Docker and tracked experiments with MLflow
Comprehensive OpenAPI documentation for easy integration

HuggingFace Docker MLflow Python DistilBERT

Code Live Demo

Fraud Detection Pipeline Optimization

PhonePe - 2024

Optimized large-scale Spark job DAG runs for fraud detection, implementing parallel processing and schema restructuring.

Reduced pipeline failures and spillovers significantly
Restructured upstream tables with optimized schemas
Created parallel flows for improved throughput
Handled millions of daily transactions

Apache Spark Airflow HiveQL YARN

LLM Risk Profile Generator

PhonePe - 2024

Fine-tuned LLM to generate real-time interactive summaries of user risk profiles for fraud investigation teams.

Real-time generation of risk summaries
Improved investigator efficiency
Integrated with existing fraud detection systems
Fine-tuned on proprietary fraud data

LLM Python NLP Risk Analysis

Technical Skills

Languages

Python SQL HiveQL Bash PySpark

ML & AI

Supervised Fine Tuning (SFT) MLflow DistilBERT NLP Transfer Learning Neural Networks and NLU CNN and RNN Machine Learning Algorithms RAG

Big Data & Tools

Apache Spark Apache Airflow YARN Hive Big Data Analytics

Data Science

Pandas NumPy Matplotlib Data Analysis Statistical Modeling SciPy Computer Vision

DevOps & Other

Docker Git QlikSense REST APIs

Domain Expertise

Fraud and Risk Detection Big Data Analysis and Modelling Data Pipeline Setup and Optimization Machine Learning and MLOPs Fine Tuning and Optimizations of LLMs