Kiran Kunapuli | Applied AI ML Lead

Applied AI ML Lead - Vice President

JPMorgan Chase

Jul 2024 - Present

Architected and delivered a multi-agent AI marketing automation assistant using LangGraph, AWS Bedrock and Claude Sonnet 4.5. Built agentic RAG workflows for semantic search, data retrieval, query reframing, feature engineering, SQL generation and modeling support, cutting campaign time-to-market from 4 weeks to 5 days.
Led the Account Engagement (CTR) model for Digital Intelligence (CCB Digital), integrating 70M+ daily transactions with customer behavior signals across credit, debit and banking activity, delivering a 35% engagement lift.

Senior Lead Data Scientist

Tokopedia

Dec 2021 - Jul 2024

Led 2 data science projects - Ads Ranking and Keyword Recommendation, managed 4 data scientists and delivered high-quality data science solutions for TopAds business problems.
Developed Real-time Ads Ranking (RTR) system, achieving 3.25% increase in CTR, 1.57% increase in Ads Revenue (~1.6 billion IDR) and deployed scalable, low latency model inference pipeline, to achieve sub 80 ms latency with 1K requests served per second.
Optimised inference of an existing deep learning solution using post-training quantization techniques which increased revenue by 1.43% (1.4 billion IDR monthly) and saved inference costs up to 166 million IDR monthly.

Data Scientist

ADP

Nov 2019 - Dec 2021

Built machine learning models for Job Classification and PayCode Classification in the ADP DataCloud team.
Implemented multi-lingual Job Classification model using Language Agnostic BERT (LaBSE) & multi-lingual Universal Sentence Encoder (mUSE), reducing time to market from 6 to 2 months.
Deployed models on AWS using Step Functions, ECR, Sagemaker Endpoint and API Gateway as batch transform pipelines and real-time endpoints, scaling model serving capacity by 3x to handle increased inference requests.

Data Scientist

DBS Bank

Oct 2017 - Oct 2019

Built machine learning models for Cash In Transit and Cheque Analytics in the Self-Service Banking, CBG group.
Implemented binary classifier using LightGBM for Cheque Risk Scoring model, to provide a risk score for all inward cheques, reduced false negatives by 18% through effective handling of class imbalance.
Forecasted no-cash situations for ATM, BTM, and CRS cash withdrawals using FB Prophet for time-series forecasting, resulting in a 15% reduction in cash shortage incidents.

Software Engineer

Tech Mahindra

Nov 2013 - Oct 2017

Worked as a Software Engineer for General Electric (GE) client in 4 different projects spanning from data ingestion to data science.
Engineered a Cash Collection ML model using clustering to segment customers based on payment behavior, decision trees to predict the likelihood of timely payments and GE US Payroll model using ARIMA.
Integrated JSON data from APIs to R dataframes and CSV, developed Talend jobs for data integration.

Building the Future with
Artificial Intelligence