Ananta Narayan Shrestha

• Building an AI-powered chatbot on AWS using Anthropic Claude and retrieval-augmented generation (RAG) to improve response grounding and quality.

AWS · Anthropic Claude · RAG

• Developed, deployed, and maintained the public-facing OpenPATH (e-mission) dashboard (Python, Docker Compose, JavaScript, MongoDB), supporting 5,000+ active users across 34 deployment programs; added dynamic configurations and inferred labels for 100% stacked bar charts, and redesigned its visualization layer, replacing pie charts with data-rich stacked bar charts.
• Architected a data validation layer using Pydantic to enforce strict schemas for incoming YAML configurations, and built a large-scale (~1TB) HPC data-processing pipeline for the EV Infrastructure Grid National Framework simulation, cutting execution time from 10 hours to 30 minutes (95%) using Polars and Parquet.
• Designed automated testing workflows (GitHub Actions, Docker, pytest) for the em-public-dashboard repo, increasing test reliability and cutting manual testing overhead by up to 60%.
• Built a machine-learning pipeline (Random Forest with feature engineering and cross-validation) for deployment-stage commute-mode inference over large-scale mobility datasets.
• Contributed to RouteE Compass (Rust routing engine enabling Google Maps' eco-routing): introduced TOML configuration injection, refactored the core crate and JSON configuration parsing, implemented an SI-unit energy model, and led conda-forge packaging for OSX arm64; recognized with the U.S. DOE-VTO AMC Team Award 2025.
• Built automated documentation workflows for the EVI-Pro project (GitHub Actions, Jupyter Book, GitHub Pages) to publish an internal documentation repo whenever a new release triggers in the private source repo.

Python · Rust · Polars · Pydantic · Docker · GitHub Actions · pytest · HPC

• Researched and integrated four fuzzing engines (Google FuzzTest, LibFuzzer, AFL++, HonggFuzz) into Yahoo's large-scale production DNS infrastructure in C++17, identifying and reporting two critical memory-management vulnerabilities.
• Engineered a 24/7 automated fuzzing pipeline and CI/CD workflows (ScrewDriver, Makefile, CMake) within a Dockerized Linux environment, and authored a ~50-page engineering report analyzing engine tradeoffs and LLVM/OpenStack integration.

C++ · LLVM · Docker · Linux · Fuzzing

• Lumify iOS App: architected a multi-threaded, low-latency (5ms) real-time audio processing library (Swift, C++, AVFoundation/AVAudioEngine) for Pulse Wave Doppler playback, collaborating with clinical scientists and meeting IEC 62304 safety-critical standards; streamlined UI development in Swift by replacing manual XIBs with programmatic StackViews, cutting new feature development time by 80% while shipping 10+ new UI controls across 10+ device sizes and orientations.
• Rhythm Service Platform (C++, MVVM): designed the UI and backend for multi-level user login using a pub-sub design pattern and layered architecture, maintaining >95% platform-layer and >85% application-layer unit test coverage.
• Women Health Care & Workflow Enhancement: shipped ultrasound-specific features (Quad AFI, High Q Doppler, 3D Preset within Preset) and a new UI/UX framework for the Blaze Toolbar supporting a dynamic array of transducers, reducing exam execution time by up to 40%.
• Triaged and resolved critical multi-threading and connectivity issues in a 20GB legacy C++ codebase; built a C++/C# static-analysis CLI tool, improving delivery speed by 30% and resolving ten high-priority production incidents.
• Designed a secure multi-level user-authentication portal (C++, C#, WPF/MVVM) with AES/SHA-256 encryption; drove end-to-end Agile feature delivery including Integration Test Plan authoring and documentation.

C++ · Swift · C# · WPF · AVFoundation · IEC 62304