Bhanu Prasanna Koppolu

Data and AI Engineer

Dortmund, DE.

About

Highly accomplished Data Scientist and AI Engineer with over two years of experience, specializing in translating complex stakeholder needs into robust, production-ready data products. Expertly co-engineered a clinical metadata platform processing 10,000+ DICOM objects daily and launched a secure analytics dashboard for radiology, leveraging deep expertise in Python, SQL, data engineering, machine learning, and AI engineering across diverse, regulated environments.

Work

Uniklinik Köln, Institute for Biomedical Informatics

|

Research Assistant - AI & Data Engineer

Köln, Germany, Germany

→

Summary

Led the productionization of clinical imaging platforms and radiology cohort-discovery applications, integrating AI/ML workflows into regulated healthcare environments.

Highlights

Productionized a clinical imaging platform from Python prototype to InterSystems IRIS for Health, implementing DICOM C-FIND/C-MOVE/C-STORE workflows across five modalities and an 11-class metadata layer, processing 10,000+ DICOM objects daily.

Developed and launched a production radiology cohort-discovery application using Dash/Plotly, FastAPI, and Keycloak, enabling role-aware metadata search and research-cohort assembly with an ingestion benchmark of 4,496 instances in 133.65 seconds.

Co-designed WavePrep, an open-source preprocessing pipeline for 22,247 ICU waveform records from 10,269 patients, reducing processing time for 1-hour/6-channel recordings to 33–51 ms with visual QA.

Developed reproducible spike-sorting benchmarks across 5 microneurography datasets (5,306 waveforms), identifying CNNs as the strongest architecture and demonstrating selective multi-task transfer benefits.

Co-designed 8 clinical cohorts for hypotension and hypoxemia prediction, quantifying distribution shift with KS/KL metrics and linking input shift to RMSE degradation (r=0.74/0.99) for a MIE 2026 study.

Education

TU Dortmund University

Dortmund, Germany, Germany
→
M.Sc. Data Science

Courses

Thesis: Stable Diffusion for BayesFlow - probabilistic diffusion models for simulation-based inference (PyTorch, Weights & Biases)

Prasad V. Potluri Siddhartha Institute of Technology

Vijayawada, India, India
→
B.Tech. Computer Science and Engineering Grade: 8.59/10

Awards

PRYSM (first place, causal AI for manufacturing quality)

Hackathon

Achieved first place for innovative contributions to causal AI in manufacturing quality during a hackathon.

VisBaltic (Audience Award)

Hackathon

Received the Audience Award at the VisBaltic hackathon, demonstrating strong project appeal and presentation skills.

Negoat.ai (Best Use of Kontext.dev)

Hackathon

Recognized for the best use of Kontext.dev at the Negoat.ai hackathon, showcasing technical proficiency with specific tools.

BONKChain (second place)

Hackathon

Secured second place at the BONKChain hackathon, highlighting competitive performance and problem-solving abilities.

Publications

MIE 2026 Distribution Shift Analysis in Generalizable Modelling: Intensive Care Time-Series Data

SHTI

Authored a publication on distribution shift analysis in generalizable modeling, specifically applied to intensive care time-series data.

Clinical time-series domain generalization study

NeurIPS

Manuscript on clinical time-series domain generalization study, currently under review for NeurIPS 2026.

MIE 2025 From Normal to Optimal: Investigating Metabolic and Inflammatory Parameters as Predictors of Survival in Locally Advanced Cervical Cancer

SHTI

Contributed to research investigating metabolic and inflammatory parameters as predictors of survival in locally advanced cervical cancer.

IEEE Stock Price Prediction using Zero-Shot Sentiment Classification

IEEE

Published research on stock price prediction leveraging zero-shot sentiment classification techniques.

IEEE Skin Lesion Classification using Transfer Learning

IEEE

Co-authored a publication focused on skin lesion classification utilizing transfer learning methodologies.

Languages

English

Certificates

AWS Academy Cloud Foundations

AWS Academy

AWS Academy Cloud Architecting

AWS Academy

Machine Learning Specialization

Andrew Ng

Mathematics for Machine Learning

Imperial College London

Skills

Data Engineering

Python, SQL, PySpark, ETL/ELT, data modelling, batch and event-driven pipelines, schema validation, data quality.

Machine Learning & Data Science

PyTorch, scikit-learn, XGBoost, time-series modelling, statistical analysis, experiment design, distribution shift, hyperparameter optimization.

AI Engineering

RAG, embeddings, hybrid retrieval, vector search, LangGraph, structured outputs, tool calling, evaluation, grounding, guardrails.

Platforms & MLOps

PostgreSQL, Databricks, Delta Lake, Airflow, Kafka/Redpanda, Redis, FastAPI, Docker, MLflow, CI/CD, AWS.

Engineering Practices

Technical documentation, testing, CI/CD, data quality, observability, reproducible workflows.

Leadership & Data Management

Data management, Reproducible data management, Git, git-annex, YODA, Workshop delivery, Stakeholder management, Cross-functional collaboration.

Projects

Allways - Multimodal Enterprise Document Intelligence Platform

→

Summary

Developed and scaled a multimodal document-intelligence pipeline for Enterprise PDFs, enabling robust data extraction, retrieval, and intelligent content processing.

Billy - Graph-Based Credit Risk ML Platform

→

Summary

Designed and deployed a graph-based machine learning platform for credit risk assessment, optimizing model training, feature engineering, and production deployment for high-throughput predictions.

Day-Ahead Electricity Price Forecasting Lakehouse

→

Summary

Transformed a university forecasting project into a robust Databricks/Delta Lake lakehouse, integrating diverse data sources for accurate day-ahead electricity price predictions.