Computer Science @ Georgia Tech | ML/Security @ GTRI CIPHER Lab | iOS Club Tech Lead

• Computing Languages and Technologies: Python, Java, C, JavaScript, C++, Swift, SQL
• Libraries/Frameworks: PyTorch, TensorFlow, scikit-learn, FAISS, NumPy, pandas, Core ML, FastAPI, Node.js
• Tools/Practices: Git/GitHub, CI/CD, Docker, Firebase, MongoDB, MVVM

Software Engineer Intern @ GTRI CIPHER Lab — Threat Analysis Branch Building ML R&D solutions and data analytic pipelines for cybersecurity and intelligence analysis workflows. Evaluating the risk posture of LLM-enabled and agentic systems within secure experimental frameworks, and applying NLP to unstructured data in a national security context.

Tech Lead @ Georgia Tech iOS Club Co-leading a 20-member project team to architect and ship production-level iOS applications with Swift, SwiftUI, and Firebase. Architecting features with MVVM, running code reviews, and mentoring developers on the team.

MARTA Live Transit Tracker — A full-stack transit tracker built to fix what the official MARTA app gets wrong. FastAPI/SQLite backend polling GTFS-Realtime and rail feeds 24/7 (575K+ arrival observations across 78 routes), a hand-written protobuf decoder validated byte-for-byte against the reference parser, and a native SwiftUI app with live vehicle tracking, delay-aware trip planning, and a home-screen widget.

LLM Red-Teaming Evaluation Pipeline — An extensible prompt-injection evaluation framework testing local and frontier LLMs against 46 attack payloads across 4 categories. Dual rule-based + LLM-as-judge scoring surfaced a robustness gap of up to 47% breakthrough on local 8B models versus near-zero on a frontier model, plus non-obvious findings like defenses that relocate vulnerabilities instead of removing them.

Multimodal AI Search Engine — A visual search engine indexing 10,000+ images with CLIP embeddings and FAISS, enabling semantically relevant image retrieval from natural language prompts with sub-2-second query times.

JotDown — A responsive iOS app built with SwiftUI and MVVM, integrating on-device Core ML for NLP-based semantic search and generating knowledge graphs to model relationships between user notes.

Real-Time ASL Interpreter — A MediaPipe + scikit-learn engineering project achieving 95% classification accuracy for static ASL alphabet gestures, with text-to-speech output and real-time webcam integration.

• Email: nmaddu3@gatech.edu
• LinkedIn: linkedin.com/in/neel-maddu
• Portfolio: neel-ai-portfolio.vercel.app