Portfolio

Rohan Isaac

Data Scientist · Atlanta, GA

Summary

I design and deploy data science solutions that transform complex data into actionable insights. Having worked across healthcare, manufacturing, CPG, and Hi-Tech, I focus on building robust, end-to-end pipelines from data curation to model deployment that support better decisions in real-world settings.

Work Experience

Emory University

Atlanta, GA

Medical Imaging AI
Data Analysis
Foundation Models
LLMs
Model Development
EHR
Data Pipeline

Data Scientist

Jul 2024 — Present

Built a data pipeline for a large-scale multi-modality breast imaging dataset of (2D and DBT mammograms, Ultrasound, MRI) for more than 1 million exams. Integrated multiple data sources such as patient workflow EHR data, risk scores, state registry data, clinical (encounters, labs, meds, notes, ICD, CPT), radiology and pathology reports extracted from EHRs.
Evaluated a commercially deployed 3D DBT breast-AI screening model, leveraging internal datasets to perform stratified analyses across outcomes, demographics, imaging characteristics and statistical evaluations to determine model optimization opportunities.
Leverage medical imaging foundation models (MammoCLIP, MedImageInsights, BiomedParse) to extract screening mammography embeddings and applied linear probing for downstream breast imaging and cancer outcome analysis.
Developed a module for structured data extraction from pathology reports by deploying open-source reasoning LLMs with a pydantic-based validation schema to extract over 30 labels spanning cancer outcomes, treatment details, and receptor status, achieving over 90% sensitivity. Applied prompt engineering techniques to improve parsing robustness and consistency.
Conducted multiple clinical analytics research for internal stakeholders specifically in breast imaging space, leveraging multiple data sources to generate data driven insights.

ClearObject

Remote

Computer vision
Edge deployment
PyTorch/TensorFlow
YOLO/U-net

Data Scientist Intern

Jan 2024 — Apr 2024

Integrated computer vision systems for manufacturing quality control, covering data collection, labeling, and model fine-tuning with PyTorch/TensorFlow (segmentation with U-net, object detection with YOLO). Built end-to-end workflows with Google Vertex AI.
Deployed models to a proprietary edge device for real-time inference using NVIDIA TensorRT and DeepStream to improve frame-rate efficiency.

Google Inc. – LibreHealth Radiology (Google Summer of Code)

Remote

Medical imaging AI
DICOM automation
Grad-CAM
Open source

Google Summer of Code Contributor

May 2023 — Aug 2023

Integrated computer vision models into LibreHealth Radiology AI for multilabel classification and object localization using Grad-CAM for chest and mammogram modalities.
Automated modality-specific model selection by reading DICOM properties (e.g., CheXpert for chest X-rays), reducing manual configuration.

Indiana University – Health Informatics Lab

Indianapolis, IN

Clinical AI study
RAG systems
Clinical EHR
Medical imaging

Research Assistant

Sep 2022 — Apr 2024

Led an NSF study on AI–Radiologist collaboration: collected and preprocessed multi-source imaging datasets, built inference scripts for chest X-rays, mammograms, and musculoskeletal X-rays, and deployed on commercial software, reducing diagnostic times by 20%.
Contributed to an open-source radiology AI system integrating imaging models for chest and breast modalities.
Built a RAG system using LangChain, Neo4j Graph DB, and ChatGPT API over clinical EHR data for semantic search and summarization.

Accenture

Bengaluru, India

Demand forecasting
Supply chain analytics
Databricks
BigQuery

Data Scientist

Apr 2021 — Jun 2022

Improved a CPG client's demand forecast accuracy by ~15% using Transformers and feature engineering; integrated into Azure Databricks workflows with MLflow experiment tracking.
Enhanced supply chain visibility via Power BI dashboards and Azure data pipelines; used Databricks (PySpark, SparkSQL), Data Factory, Synapse, and Tabular models.

Senior Analytics Consultant

Apr 2020 — Mar 2021

Improved home appliances demand forecasts by 5% over baseline using an ensemble of tree-based models on Google Cloud with tailored feature engineering.
Built and scheduled complex SQL queries on BigQuery through Informatica ETL pipelines to move data from lakes to marts.

Analytics Consultant

Apr 2019 — Mar 2020

Built a sourcing optimization solution using text analytics and NLP similarity matching, improving production efficiency by 20% for a CPG client.
Developed large-scale data preprocessing pipelines (SparkSQL, PySpark, Pandas) on Azure to build a supply forecasting solution.

Tata Consultancy Services

Chennai, India

Oracle ERP
SQL reporting
Enterprise support

Systems Engineer

Oct 2016 — Jun 2018

Managed Oracle ERP service requests and built complex SQL queries for custom reports for a high-tech client.

Education

Master of Science, Applied Data Science

Indiana University-Indianapolis

Aug 2022 — May 2024

Post Graduate Program Data Science

Praxis Business School

Jul 2018 — Mar 2019

Bachelor of Engineering, Mechanical

Anna University

Oct 2012 — Jun 2016

Skills

Programming & AI

Python
R
SQL
PyTorch
TensorFlow

Cloud Tools

Google Cloud Platform (Cloud Run, Cloud Functions, BigQuery, VertexAI, Firebase)
Azure (Databricks, Synapse, DataFactory)

Domain Expertise

Medical imaging AI
Healthcare Analytics
Supply Chain Analytics

Publications

Abstracts

Interests

AI and Gen AI
Health Informatics
Tech Trends
Travel
Cooking

Rohan Isaac

Summary

Work Experience

Emory University

Data Scientist

ClearObject

Data Scientist Intern

Google Inc. – LibreHealth Radiology (Google Summer of Code)

Google Summer of Code Contributor

Indiana University – Health Informatics Lab

Research Assistant

Accenture

Data Scientist

Senior Analytics Consultant

Analytics Consultant

Tata Consultancy Services

Systems Engineer

Education

Master of Science, Applied Data Science

Post Graduate Program Data Science

Bachelor of Engineering, Mechanical

Skills

Programming & AI

Cloud Tools

Domain Expertise

Subgroup Performance of a Commercial Digital Breast Tomosynthesis Model for Breast Cancer Detection

Impact of label noise from large language models generated annotations on evaluation of diagnostic model performance

Evaluating Vision Language Models (VLMs) for Radiology: A Comprehensive Analysis

Novel AI-Based Quantification of Breast Arterial Calcification to Predict Cardiovascular Risk

Real-world performance evaluation of a commercial deep learning model for intracranial hemorrhage detection

Feature Quality and Adaptability of Medical Foundation Models: A Comparative Evaluation for Radiographic Classification and Segmentation

Measuring impact of radiologist-AI collaboration: Efficiency, accuracy, and clinical impact

A general-purpose AI assistant embedded in an open-source radiology information system

Comparing General Vision Transformer Embeddings versus Task-Specific Deep Learning Models for Knee Osteoarthritis Grading

Leveraging Pre-trained Medical Image Embeddings for Knee Pain Score Prediction: A Comparative Analysis of Vision Transformer and CNN Approaches

Systematic Detection and Correction of DICOM Label Discrepancies in Large-Scale Chest X-ray Datasets

On the Feasibility of Chest X-ray Reconstruction from Foundation Model Vector Embeddings

Report Interpretation Times for Radiology Residents and Attendings Before and After Deployment of a Triaging Model for Pulmonary Embolism and Intracranial Hemorrhage Detection in a Large Academic Center

LLM Showdown: Benchmarking Performance, Cost, and Speed of Granular Ground Truth Extraction in Radiology Reports for Post-Deployment Monitoring of AI Models

Chest Radiography and Mammography-Derived Imaging Biomarkers Enhance Cardiovascular Risk Prediction Beyond ASCVD Risk Score

EdgeDeID: Advancing De-identification with Small LMs (SLMs) and Synthetic Data on Edge Devices

Emory Breast Imaging Dataset (EMBED) v2 – A racially diverse, multi-modal dataset of 1.2M breast imaging exams and associated histopathology

Score Change as a Marker of Elevated Risk in AI-Based Breast Cancer Screening

Association Between False Positive Mammograms and Return to Screening in a Racially Diverse Cohort

Interests