Raaj Atulkumar Patel

• ●Problem: Financial queries need trustworthy, non-hallucinated answers grounded in real data.
• ●Built: Full-stack system converting natural language -> SQL -> retrieved context -> grounded explanations.
• ●Built: RAG pipeline where all numerical values are strictly fetched from SQLite (LLM used only for reasoning).
• ●Built: Provider-swappable LLM architecture (Ollama local, OpenAI, Groq, HuggingFace, vLLM).
• ●Built: CRUD APIs, analytics SQL, budget vs actual analysis, anomaly detection, plus schema/index tuning.

• ●Built: End-to-end DER simulations combining electrical telemetry and cyber events.
• ●Built: Modeled false data injection, spoofing, and command manipulation attacks.
• ●Built: Generated ML-ready labeled datasets for anomaly and intrusion detection.

• ●Built: Evaluated lightweight open-source LLMs to learn normal vs anomalous telemetry patterns.
• ●Built: Prompt-based classification optimized for low-latency inference and rapid triage.

• ●Built: End-to-end computer vision pipeline using PyTorch and OpenCV.
• ●Built: Applied preprocessing, augmentation, training, and evaluation.

Purdue University Northwest | Advisor: Prof. Shafkat Islam

• ●Designing AI-driven detection pipelines for cyberattacks in Distributed Energy Resource (DER) systems.
• ●Building OpenDSS-based IEEE 13-bus and 123-bus simulations with integrated PV/BESS assets for controlled anomaly studies.
• ●Creating reproducible datasets and evaluation pipelines to support research-grade ML experimentation and reporting.

Business Assistant & Technical Contributor

• ●Built and deployed the company website (navarang.com) and introduced lightweight digital tooling to improve operations.

Engineering Intern

• ●Supported AI and embedded workflows via Linux-based scripting, debugging, and structured testing.