← Return to OverviewVerified Implementations

Engineered Systems & Case Studies

Each project represents a verifiable system built under empirical rigor—bridging hardware microcontrollers, econometric modeling, and production-ready retrieval pipelines.

Receiving the Top Innovator award on stage from Arduino CEO Fabio Violante for our project, SafeGuard-AI.

National Winner — Qualcomm AI for AllJan 2026 – Feb 2026

SafeGuard-AI: TinyML Fall Detection System

Real-time on-device fall detection deployed on Arduino UNO Q with quantized inference

A TinyML physical AI system engineered for ultra-low latency fall detection. Designed and trained a deep learning model using IMU sensor streams and engineered motion features, quantized with TensorFlow Lite, and deployed for real-time on-device inference on an Arduino UNO Q.

TinyMLTensorFlow LiteQuantizationArduino UNO QIMU SensorsMotion Feature EngineeringOn-Device Inference

Read Full Case Study Arduino Project Hub GitHub Code

Research team at the 3rd International Conference on Infrastructure Development & Sustainability (ICIDS 2025)

First-Author Paper — ICIDS 2025Jul 2025 – Dec 2025

Analyzing Indian Stock Markets Through Correlations: Comparative Insights

First-author empirical research modeling systemic risk and lead–lag dynamics across Indian market sectors

Comparative analysis evaluating dependency measures across Indian equity sectors using empirical time-series correlation matrices.

Pearson CorrelationSpearman RankKendall's TauTime-Lagged Cross-CorrelationMinimum Spanning TreesSystemic RiskTime-Series Preprocessing

Read Full Case Study Publisher DOI

SYSTEM ARCHITECTURE

High-Level System Flow

Academic Content→Retrieval→Context LLM

Academic CollaborationJun 2026 – Present

Moodle AI Assistant

RAG-Based University AI Platform

Developing a Retrieval-Augmented Generation (RAG) based assistant for university academic content, enabling students to interact with course material through context-aware AI responses as part of a broader academic AI initiative.

RAGLLMsAcademic RetrievalContext-Aware QAOpen-Source ModelsModel Comparison

Read Full Case Study

SYSTEM ARCHITECTURE

Multi-Layered Anomaly Engine

Rules Filter→LLM Signals→Audit Review

Anomaly Detection & MLApr 2024 – Jul 2024

Hybrid Fraud Detection System

Multi-layered anomaly detection combining deterministic heuristic rules and LLM-based signals

A financial anomaly detection pipeline integrating rule-based transaction validation with LLM reasoning signals to flag suspicious financial transactions, suppress false positives, and route ambiguous cases to a human-in-the-loop review workflow.

Hybrid PipelineHeuristic RulesLLM SignalsAnomaly DetectionHuman-in-the-LoopFinancial Risk

Read Full Case Study GitHub Profile