Saloni Shah

• ▹Leveraged AI to build a web-based codebase exploration tool tailored for Project Management, Systems, and SW Development teams to navigate a large, undocumented C++ codebase, providing natural language chat, multi-repo indexing across 6 languages, dependency graphs, and coupling analysis.
• ▹Enabled unattended Ubuntu 24.04 network installs in an air-gapped enclave by authoring the autoinstall config and supporting scripts from scratch, making it possible to provision new servers without manual intervention.
• ▹Set up bare-metal NFS storage for a defense software program by curating an NFS server and integrating it into a Kubernetes cluster using Ansible and standard infrastructure design patterns.
• ▹Extended an internal MagicDraw plugin to allow engineers to update architecture models directly through Claude Code, supporting model element updates, diagram creation, and a validation summary mode that eliminated 80% of token usage in a full modeling session.

• ▹Reduced trace decoding runtime by 9x by replacing a Python decoder with a C Babeltrace 2 plugin for inline IPv4/IPv6/MAC address decoding, accelerating large-scale networking debugging workflows.
• ▹Enhanced IS-IS, RIB, and FIB modules by adding last-update-time and failure reason attributes, standardizing YANG models and improving root-cause analysis of routing delays.
• ▹Designed, tested, and shipped enhancements across ArcOS, reducing CPU overhead and strengthening reliability in large-scale enterprise deployments.

• ▹Led weekly lab sessions for 14 students across core electrical and computer engineering courses (EE 1301 & EE 2301), supporting hands-on experimentation and problem-solving.
• ▹Guided students through circuit concepts, debugging techniques, and lab instrumentation while reinforcing lecture material.

• ▹Built an AI scheduling app that converts natural language task input into optimized Google Calendar events using LLM parsing to extract tasks, estimate durations, and auto-schedule into free time slots.
• ▹Developed full-stack application with Next.js, Supabase, and OpenAI API, integrating Whisper for voice input and deploying as a PWA on Vercel. Currently in active beta with real users.

• ▹Built a prototype for Meta smart glasses integrating Google Gemini multimodal models for real-time vision and audio context processing.
• ▹Implemented a multimodal inference pipeline combining computer vision, speech analysis, and LLM reasoning to identify individuals, extract conversational updates, and generate post-interaction artifacts using NanoBanana.

• ▹Developed a Model Context Protocol (MCP) server enabling AI assistants to integrate with Google Classroom API for educational workflow automation.
• ▹Configured OAuth2 authentication with Google Cloud Platform, managing secure token storage and refresh mechanisms.
• ▹Architected comprehensive TypeScript API wrapper supporting 8+ operations including course management, assignment creation, and submission monitoring.

• ▹Implemented fully playable Flappy Bird on an 8×8 LED matrix with tilt and shake gesture controls via I2C accelerometer, writing three embedded C libraries from scratch.
• ▹Architected a three-state game engine with independent timers, collision detection, and a linear congruential generator for randomized pipe generation.

• ▹Architected comprehensive programming curriculum teaching software development fundamentals to K-8 students.

• ▹Orchestrated a team of 20 members in designing and fabricating competition-ready robot parts using Fusion 360 and CNC tools.

Activities