Ravi Chowdhary

Data Engineering & Machine Learning

My background spans nearly five years of experience in enterprise healthcare data systems and analytics, along with academic training in biomedical engineering, computer science, and artificial intelligence.

I'm interested in building reliable data systems, understanding how machine learning works, and exploring how these technologies can be applied to challenging real-world problems.

Explore selected work LinkedIn

Ravi Chowdhary seated outdoors among autumn flowers

SELECTED FOCUS AREASPYTHON / SQL / PYSPARKPYTORCH / ONNX / TENSORRTAZURE / DATABRICKS / POWER BI

01 — THE WORK

Selected projects.

Selected studies in transformer inference, document retrieval, and graph machine learning, with an emphasis on experimental evaluation, reproducibility, and understanding technical limitations.

PROJECT 01 / FEATUREDSource private

TRANSFORMER INFERENCE · PERFORMANCE STUDY

DistilBERT · SST-2 Sentiment classification

CPU INFERENCE PyTorch ONNX Runtime

GPU INFERENCE NVIDIA TensorRT

VALIDATION & BENCHMARKING Numerical agreement · Latency · Throughput

INFERENCE ENGINEERING

Transformer Inference Performance Lab

An experimental study of DistilBERT sentiment-classification inference using PyTorch and ONNX Runtime on CPU and NVIDIA TensorRT on an L4 GPU. Examined numerical agreement, latency, and throughput across 12 batch-size and sequence-length configurations, including FP32, INT8, and mixed-FP16 experiments.

12 workload configurations362 passing automated tests

PyTorchONNX RuntimeTensorRTQuantization

PROJECT 02Source private

RETRIEVAL EVALUATION

RAG Document Intelligence Evaluation

A reproducible evaluation of embedding-based document retrieval using five source documents and 10 evaluation questions. Examined retrieval results, evidence selection, and failure modes to understand when relevant information was successfully retrieved and when retrieval failed.

10 evaluation questions5 source documents

EmbeddingsRetrievalFailure Analysis

PROJECT 03Source private

GRAPH MACHINE LEARNING

Graph Learning on Citation Networks

An experimental comparison of GraphSAGE and an MLP baseline for classifying research papers in the ogbn-arxiv citation network. Examined prediction accuracy, neighbor sampling, and resource-bounded batched evaluation on CPU.

53.7% GraphSAGE test accuracy49.9% MLP test accuracy

GraphSAGEPyTorch GeometricOGB

Development approach. These projects were developed with substantial ChatGPT assistance, including technical guidance, implementation instructions, and troubleshooting. I carried out the guided implementation steps, ran experiments and tests, examined the results, and documented the findings.

Project source repositories are currently private. Any future publication of source code or technical documentation will depend on separate technical, attribution, and licensing reviews. Project metrics reflect recorded experimental results; diagrams are conceptual illustrations.

02 — PROFESSIONAL EXPERIENCE

Enterprise experience.
Data & analytics.

At AmeriHealth Caritas, I spent nearly five years working with healthcare data systems, using SQL and PySpark for data analysis, supporting migrations and integrations, and developing analytical reports.

AmeriHealth Caritas

Business Systems Analyst I → II

2021 — 2026

Analyzed legacy SAS and Scala data-processing logic and supported Hadoop-to-Azure migration using SQL, PySpark, and Azure Databricks.
Evaluated data ingestion and integration workflows through proof-of-concept testing, source-to-target validation, and data reconciliation.
Used SQL to check data quality and developed Power BI reports to support healthcare data analysis.
Supported secure API integration proofs of concept involving Apigee, Azure Application Gateway, and mutual TLS (mTLS).

SQL · PySpark · Azure Databricks · Azure Data Factory · Power BI

03 — ABOUT & LEARNING

Interested in data, models,
and how systems work.

I'm interested in how data moves between systems, how it can be analyzed and made useful, and how data systems support machine-learning applications. I also seek to understand how machine-learning models perform under real constraints. My interests include data engineering and analytics, ML infrastructure, inference performance, graph learning, and AI in healthcare and biomedicine.

I value reproducibility, careful attribution, and drawing conclusions that the evidence supports. I continue to strengthen my understanding of the code and methods behind my projects through review, testing, and technical study.

EDUCATION

Georgia Institute of TechnologyM.S. in Computer Science · AI specializationAll requirements fulfilled; degree expected December 2026

University of MinnesotaM.S. in Biomedical Engineering

Duke UniversityM.S. in Biomedical Engineering

University of Illinois Urbana-ChampaignB.S. in Bioengineering

04 — COLLABORATIVE ACADEMIC WORK

AI for ovarian cancer subtype classification.

In a team project using the UBC-OCEAN histology dataset, I participated in the comparison of deep-learning architectures for ovarian cancer subtype classification, helped coordinate experiments, and compiled results. The team's fine-tuned ResNet26 achieved a reported test F1 score of 0.50.

Computer visionDigital pathologyModel comparison

05 — CONNECT

Interested in the work?
Let's talk.

I'm interested in connecting with people working in data engineering, analytics, applied AI, and machine learning systems. Whether you'd like to discuss technical ideas, exchange perspectives, or explore a professional opportunity, I'd be glad to hear from you. You can also find me on LinkedIn and visit my GitHub profile.

Connect on LinkedIn GitHub profile

The repositories for the three featured projects are currently private.