Data Scientist and MS Computer Science student building predictive models, production ML pipelines, and stakeholder-facing analytics from large, multi-source datasets. My background spans physics-simulation surrogate modeling, procurement analytics, and hyperspectral imaging research — with 2 patents and 3 publications along the way.

• 🔭 Currently a Research Assistant at the University of Cincinnati, building PyTorch/scikit-learn surrogate models for large-scale physics simulations
• 💼 Previously Data Scientist & Engineer at Infosys, shipping ETL pipelines and GPT-4/LangChain extraction workflows across 30+ data sources
• 🔬 3 publications and 2 patents in hyperspectral imaging and ML
• 🌱 Currently deepening MLOps and GenAI/RAG systems work
• 📫 Reach me at muddanvk@mail.uc.edu or LinkedIn

• Surrogate modeling for physics simulations (University of Cincinnati) — PyTorch/scikit-learn models that replace 8-hour simulation runs with sub-second predictions, screening 85% of candidate designs before full analysis.
• Procurement & vendor analytics at scale (Infosys) — ETL pipelines on AWS S3/Parquet ingesting 30+ APIs and file sources, plus GPT-4/LangChain extraction workflows that turn unstructured procurement documents into structured, versioned features.
• Hyperspectral image classification (Hispec Lab, SRM University AP) — a production HSI pipeline on EO-1 Hyperion satellite imagery (ENVI correction → endmember detection); benchmarked 25 deep learning and ensemble architectures, with a Stacking LSTM-CNN model outperforming all standalone baselines.
• Self-serve analytics platform (NorthPeak) — a dbt/Kimball warehouse over 3.3M+ e-commerce records with a governed metric dictionary, 73 dbt tests, 25 Great Expectations checks, and Dagster orchestration.
• Causal marketing attribution (Streamly) — Shapley-value multi-touch attribution validated against a known ground truth, recovering true channel importance with 82% less error than last-touch.

• Muddana, V.K.S., et al. "Efficient Hyperspectral Image Classification of the Krishna River Basin in Andhra Pradesh Using Hybrid Ensemble Learning Models." Optica Imaging Congress, 2025. Link
• Muddana, V.K.S., et al. "Hyperspectral Image Classification with Deep Learning: Unleashed by Feature Selection and Extraction." Innovations in Computer Science and Engineering, Springer Nature, 2025. Link
• Muddana, V.K.S., et al. "Non-Invasive Oral Cancer Detection Using Hyperspectral Imaging and Advanced Spectral Unmixing Models." IEEE, 2024. Link