My Projects

A full-stack multi-agent mentoring platform currently used by research students at Ozlab, Nagoya Institute of Technology, for weekly progress report feedback. The system generates structured feedback through a Report Analyst Agent and includes an Alignment Agent that lets students challenge AI feedback with evidence-based justifications, with a supervisor-in-the-loop adaptive memory that encodes supervisor resolutions as reusable rules.

A multi-agent system for autonomous student risk identification, voice-AI outreach for intervention, and real-time human-in-the-loop workflows for supervisor oversight, supporting multiple concurrent users. The system monitors student progress signals, triages risk cases through automated routing, conducts voice-based check-ins with flagged students, and escalates to supervisors with conversation transcripts and analytics.

Built an intelligent blockchain analytics platform that transforms complex Solana DeFi data into accessible insights through natural language queries. Combines advanced AI with real-time market analysis to revolutionize trader interactions with the Solana ecosystem.

Investigated and analyzed absenteeism trends in a medium-sized company using Power BI to identify insights and drive organizational performance improvements.

Identified irregularities in the general ledger of a large company using rule-based analysis with Pandas, focusing on journal entries that do not net to zero.

Implemented an anomaly detection algorithm using Gaussian distribution to detect anomalous behavior in servers on a network, preventing system failures and improving performance.

Built two deep learning models for diabetic retinopathy detection: a binary-class ResNet-50 model (85,056 fundus images) and a multi-class ResNet-18 model (44,224 fundus images).

Led a team of 20 interns in developing a dementia prediction model using brain MRI scans. Analyzed a tabular dataset of 150 subjects aged 60-96 years.