Harshavardhan Shahi

Washington University in St. Louis

Advisor: Dr. Netanel Raviv

• Extended random-linear-code encodings for Vector Symbolic Architectures with concatenated Reed-Solomon–Hadamard codes, achieving provable quasi-orthogonality and noise-resilient recovery.
• Implemented and benchmarked binding and bundling recovery algorithms in Python; prototyped integration with the TorchHD library.

Advisor: Dr. Kim Taeho — School of Medicine, Center for Advanced Medicine

• Developed optical flow computer vision models to correct MRI geometric distortion for radiation therapy planning, reducing distortion error by 1%.

MindMirror, Truman State University

• Trained neural networks in TensorFlow on self-collected EEG data via OpenBCI to classify hearing signals in the brain.

MilliporeSigma, St. Louis, MO

• Built two Model Context Protocol servers exposing internal data sources, including Aldrich Market Select compound search and checkout, to LLM-based research agents.
• Designed typed data contracts and OAuth authentication for each server, enabling secure access for public use.
• Reduced the time for researchers to locate and order compounds from ten minutes to twenty seconds by replacing manual lookup with agent-driven workflows.
• Won Best Intern Project and Presentation out of forty interns.

Centene Corporation, St. Louis, MO

• Deployed 50+ production ML models across Clinical, Finance and Quality lines of business, serving 25 million members.
• Built CI/CD pipelines in GitLab to containerize models with Docker and deploy to Kubernetes, cutting release time from forty minutes to two.
• Developed models in XGBoost, PyTorch and spaCy on Databricks, sourcing data from Snowflake and Teradata.
• Implemented runtime model monitoring with MLflow, tracking metrics across 100+ models.
• Reduced inference latency by up to 50% on several models by optimizing steps of the inference pipeline.
• Built executive dashboards in Python and R reporting model and company performance to C-suite leadership.
• Won first place twice in internal hackathons: a graphical patient information viewer with RAG synopsis (2024), and a site consolidating similar dashboards to reduce redundancy (2023).

Centene Corporation, Kirksville, MO

• Built and deployed a behavioral health prediction model to Kubernetes via Docker, returning over $1.1 million.
• Modeled customer lifetime value to support the care manager and financial teams.
• Hired into the full-time ML Engineer role upon graduation.

M.S. Computer Science, machine learning focus · GPA 3.89

B.S. Computer Science & Mathematics, minor in Statistics · GPA 3.74 · Department Honors