Hi, I'm Vibhu.

1. 01

   EvalWard LLM regression gate for CI/CD

   Aug 2026

   Personal project · open source

   A CI gate that catches silent quality regressions in LLM-powered features. It re-runs a bank complaint-triage prompt against a 150-case golden dataset built from the public CFPB Consumer Complaint Database, diffs macro-F1 against the last promoted baseline, and blocks the merge when quality drops materially — naming the exact cases that flipped.

   Python ·Anthropic Claude API ·DeepEval ·Pydantic ·Typer ·SQLite ·Streamlit ·Docker ·GitHub Actions

   View on GitHub
2. 02

   MCP Tutor Agent

   Jul 2026

   Agentic AI coursework

   A local-first AI tutoring agent that calls external tools over the Model Context Protocol. Two MCP servers in C# (stdio and streamable HTTP transports) feed their tool catalogs to an LLM agent via the Microsoft Agent Framework, with a streaming Blazor chat UI orchestrated by .NET Aspire.

   C# ·.NET Aspire ·MCP ·Microsoft Agent Framework ·Blazor ·Ollama

   View on GitHub
3. 03

   NC Early Mathematics Placement Testing Online Testing System

   Aug 2025 — May 2026

   Senior Capstone · 5-person Agile team

   Full-stack web platform used by 60 NC high schools and 5,000+ students. Role-based access with JWT + 2FA, real-time proctor monitoring over SignalR, and an async email pipeline on GCP Pub/Sub. Shipped through GitHub Actions CI/CD.

   React 19 ·Vite ·.NET 8 ·EF Core ·PostgreSQL ·SignalR ·JWT ·GCP ·GitHub Actions
4. 04

   Metamorphic Testing of Multi-Agent AI Systems

   Apr 2026

   Research

   A novel catalog of 9 metamorphic relations for testing multi-agent LLM systems where traditional oracles fail — interchangeability, redundancy invariance, graceful degradation. Reusable Python framework with semantic verifiers; CrewAI harnesses across OpenAI and Ollama via LangChain.

   Python ·CrewAI ·LangChain ·OpenAI ·Ollama
5. 05

   Dental Charting Tool

   Jun 2025 — Jan 2026

   Applied research · ECU School of Dental Medicine

   WPF/.NET desktop app for radiograph analysis with GPT-4o Vision, FDI tooth-numbering, CDT-coded procedure cataloging, and urgency triage. A 5-step prompt pipeline keeps output clinically contextualized. Presented at ECU RCAW and CORAS 2026.

   C# ·WPF ·.NET ·GPT-4o Vision ·Ollama

   View poster (CORAS 2026)

Experience

1. Jun 2025 — May 2026

   Undergraduate Research Assistant

   ECU College of Engineering and Technology

   Built data ingestion prototypes in Python, C#, and Java across three faculty research projects in health tech and cybersecurity. Maintained Python/Maven environments and local LLM integrations supporting team-wide reproducibility.

2. Jan 2023 — Jan 2026

   Learning Assistant

   Thomas Harriot College of Arts & Sciences

   Tutored 50+ students in Algebra, Calculus, and Statistics through weekly review sessions and office hours — raising failing test scores to 80%+ and helping 75% of students earn a 4.0 in the course.

Stack

AI / ML

Agentic AI Systems ·Multi-Agent Systems ·LLM Application Development ·Model Context Protocol (MCP) ·Microsoft Agent Framework ·CrewAI ·LangChain ·Prompt Engineering ·LLM Evaluation & Testing ·LLM-as-a-Judge ·DeepEval ·Golden Datasets & Regression Gating ·Anthropic Claude API ·OpenAI / GPT-4o Vision ·Ollama ·PyTorch ·NLP

Languages

Python ·C# ·Rust ·Java ·JavaScript ·TypeScript ·SQL

Frameworks & Web

.NET Aspire ·.NET 8 ·ASP.NET Core ·Blazor ·Entity Framework Core ·WPF ·React ·Next.js ·Vite ·SignalR ·REST APIs ·JWT ·Pydantic ·Typer ·Streamlit

Cloud & Tools

GCP (App Engine, Cloud Run) ·Pub/Sub ·Docker ·GitHub Actions ·GitLab CI ·PostgreSQL ·SQLite ·Git ·Bash ·PowerShell ·pytest ·Ruff ·mypy