---
title: "Zenith AI — Parth Kapoor"
description: "AI hiring assistant that scores, retrieves and ranks candidates"
canonical: "https://parthkapoor.me/builds/zenith-ai"
---

# Zenith AI

> AI hiring assistant that scores, retrieves and ranks candidates

Zenith AI is a recruiting tool for teams screening candidates against a job description: recruiters describe a role in plain language to a chatbot, which turns that into a structured job-requirements schema and returns a ranked shortlist pulled from a vector database of candidate profiles.

A LangChain-driven chatbot backed by an Ollama LLM holds conversational memory while it extracts and validates job requirements against a Pydantic schema; candidate resumes are embedded with Hugging Face sentence-transformers and stored in Upstash Vector for similarity search. The FastAPI Python service sits behind a Next.js/Drizzle/MySQL frontend where candidates create accounts and submit resumes.

- Status: shipped
- Language: TypeScript
- Started: Dec 2024
- Shipped: Mar 2025
- Live: https://zenith.parthkapoor.me
- Source: https://github.com/parthkapoor-dev/zenith.ai

## Highlights

- 78%+ ranking accuracy across 100+ resumes via RAG plus vector search
- Cut recruiter screening time by 70% versus manual review
- Chatbot converts free-text hiring queries into a structured JobRequirements schema before searching
- Candidate profiles embedded via sentence-transformers and stored in Upstash Vector for similarity matching
- Next.js/Drizzle/MySQL frontend paired with a self-hostable FastAPI + LangChain backend

## Stack

- **Frontend**: Next.js, Drizzle ORM, MySQL, Tailwind CSS
- **Backend**: FastAPI, LangChain, Ollama
- **AI/ML**: Hugging Face Transformers, Sentence-Transformers, Upstash Vector

All builds: [https://parthkapoor.me/builds](https://parthkapoor.me/builds.md)