Resume

Summary

As a research engineer with 14 years of experience, I have experience in complex domains, from medical computer vision to realtime entertainment. My passion for continued learning fuels my ability to innovate, as evidenced by projects like developing a 3D U-Net architecture for segmenting CT scans and securing FDA Class II certification for a computer vision system used to view patient-specific fractures.

Recent Experience

• Developed a deep learning pipeline for segmenting bones from CT scans with over 100,000,000 voxels using off-the-shelf GPUs. This project saved time and effort needed to research and develop complex hole-filling algorithms. I created a 3D U-Net model using Keras and TensorFlow.
• Researched and developed automatic and semi-interactive methods for identifying fracture pieces. This project used traditional computer vision and relied on tools like ITK, OpenCV, scikit-image, MATLAB, and the Python data science stack. I then integrated this work into the Unreal Engine for VR using C++ and Blueprints.
• Architected and implemented a secure pipeline for creating patient-specific 3D meshes based on CT scans. Designed a bench test and wrote a report demonstrating the system’s accuracy, which helped the product obtain its Class II medical device approval from the FDA.
• Used Houdini, a tool used in the VFX industry, to create 3D models used for exploratory data analysis. This facilitated stakeholder conversations, as looking at a 3D model was easier to understand than discussing a Sørensen-Dice Coefficient.
• Managed a team responsible for building the platform used to collect, store, and analyze data vital to the company. The team consisted of three in-house engineers and two contract engineers.
• Migrated a legacy tech stack to a type-safe architecture using TypeScript, Zod, React, and Vite through a gradual integration. Architected a data-driven system for adding new products to the data collection platform, which reduced developers’time by over 75%.

• Eliminated an entire category of bugs and allowed game designers to update game logic without a programmer. Collaborated with a game designer to build a single source of truth for the game’s core systems by introducing Lua into the client and the server. I integrated Lua into the server.
• Developed achievement and event systems used to increase player retention using Scala and Akka. Worked with the Unity client to integrate the systems.
• Advocated for the use of bots to test the server. My efforts reduced server developer reticence. The bots helped discover more bugs and allowed designers to test gameplay ideas quicker with no backend server developer time required.

• Wrote an artificial intelligence and ability system that let game designers experiment with agent behavior with minimal engineer intervention. The agents used a finite-state machine architecture and eventually moved to a behavior tree system.
• Created an image management system using Unity, Angular, and C# on both the client and the server. This packed customer assets into an efficient texture atlas, which reduced per-asset network calls and storage.
• Worked on an augmented reality application using Google’s Project Tango. Used a combination of native Android development and Unity.

Highlighted Skills

Languages

Python, C++, TypeScript, JavaScript, Scala, SQL

Computer Vision

OpenCV, skimage, ITK

AI/ML/Data

Keras, TensorFlow, PyTorch, 3D deep learning, LangChain, NumPy, HDF5, Pandas, scikit-learn, matplotlib, Jupyter

Deployment

Docker, GitHub Actions, Azure

Management

Jira, 1-on-1 meetings, code reviews, mentoring, interviewing

Analytical

Linear Algebra, Algorithm Analysis, Single and Multivariate Calculus

Education

• Member of the Secure Wireless Ad hoc Network Laboratory

• Awarded Computer Science Student of the Year sophomore and junior years.

Publications

Sirisha Medidi and Peter Cappetto, ”History-based Route Selection for Mobile Ad Hoc Networks”, In Proceedings of 15th IEEE International Conference on Networks ICON2007, November 2007, Adelaide, South Australia.

Sirisha Medidi and Peter Cappetto, ”History-based Route Selection for Reactive Ad Hoc Routing Protocols”, In Proceedings of the SPIE, May, 2007, Orlando, Florida.