ResuMate AI：7 个 AI 智能体驱动的全栈招聘平台

ResuMate AI

A full-stack multi-agent AI hiring platform with real-time avatar interviews, career coaching, intelligent candidate evaluation, and an autonomous AI hiring agent that runs your entire pipeline through voice conversation.

ResuMate AI isn't a wrapper around ChatGPT — it's a production-grade system where 7 specialized AI agents orchestrate the entire hiring pipeline, from resume parsing to live video interviews with a lip-synced AI avatar — all controllable through Jarvis, a conversational AI that automates hiring end-to-end.

Demo

Live Demo: resumate-ui.onrender.com

Free-tier hosting — first load may take ~30s to wake up.

Platform Preview

Landing Page & Agent Showcase

Hiring Manager Dashboard

Upload & Analytics

Upload resumes, view analytics, and let AI rank candidates automatically.

AI Chat & Candidate Focus

Multi-candidate AI chat, deep-dive candidate focus with GitHub scanning and resume intelligence.

Candidate Evaluation

AI-powered evaluation reports with role fit scoring, strengths, and growth areas.

Interview Setup & Scheduling

Create AI interviews, draft professional emails, and schedule meetings — all from one place.

Candidate Portal

Login, Dashboard & Analytics

Email-based authentication, resume analysis, and personalized analytics.

AI Advisor & Live Interview

AI career advisor with resume coaching, interview prep — then step into a live video interview with an AI avatar interviewer featuring face tracking and proctoring.

Architecture

System Overview

Agent Framework

Features

Jarvis — AI Hiring Copilot

Jarvis is a conversational AI hiring chief-of-staff that turns the hiring manager dashboard into one continuous voice/text workflow. It sits on top of the existing ATS, evaluation, enrichment, interview, reporting, and export features and orchestrates them through natural conversation.

Talk to Jarvis like a colleague:

• "Screen everyone for a Python Developer role" → Jarvis runs ATS scoring on all uploaded resumes and names the top candidates with scores
• "Check his GitHub" → live GitHub profile analysis: repos, languages, top projects, AI technical impression
• "Give me his drawbacks" → full AI evaluation report with specific Growth Areas, fit score, and strengths
• "Show me her resume red flags" → resume intelligence with confidence score, gaps, verification targets, and red flags
• "Set up interviews for the top 3" → creates interviews and shows a ready-to-send email draft with a single Send button
• "Create an interview for Maya" → builds a single-candidate interview with role, level, focus areas, and candidate access
• "How did his interview go?" → fetches the completed interview report with scores, violations, eye contact, and summary
• "Did he exaggerate on the resume?" → runs credibility analysis by comparing resume claims against interview performance
• "Open the PDF report" → launches the branded downloadable assessment report
• "What's the market rate for this role in Bangalore?" → live web search with sourced findings
• "Find his LinkedIn" → full profile enrichment: GitHub, LinkedIn, contact info, AI summary
• "Share my Calendly link" → fetches your scheduling link to send to candidates

Jarvis keeps session state for the active role, active candidate, shortlisted candidates, interruption state, and cached artifacts such as GitHub analysis, resume intelligence, evaluations, interview configs, interview reports, credibility checks, and research notes. That lets follow-ups like "check his GitHub", "give me her drawbacks", or "send it" resolve naturally without restating context.

Every result appears as an interactive card in the chat — expand any card for full untruncated detail. Jarvis keeps spoken replies short and human while the cards expose the full operational output from the underlying agents and APIs.

Find Candidates — Talent Mapping for the Entire Pool

Describe who you want in plain words — "Backend engineer in Boston, strong Python, has shipped a real product; ex-founder is a plus". The Sourcer Agent turns that into a handful of must-have and nice-to-have criteria, then reads everyone it can reach — your uploaded resumes, GitHub, and public profiles found by web search — and writes a judgement on every single person, instead of keyword-filtering first and reading only what survives.

• Live talent map — screened, judgements written, shortlisted, people per second, elapsed and cost so far; a live sample of the people being read; and a population map with one cell per person
• Judging now — for the person being read: what their profile shows for each criterion, how strongly, and the written judgement
• Why filters would have missed them — every shortlisted person is checked against the title + keyword search a recruiter would have run, so the page shows Boolean + filters: N vs Read everyone: +M, and why the filter missed them (non-standard title, career pivot, no brand-name employer, thin profile)
• Scores you can check — the model only rates each criterion; the score is fixed arithmetic (must-haves weigh double, a clearly missed must-have keeps someone off the shortlist), so the same answers always give the same number
• Human in the loop — save, dismiss, and draft outreach that cites the person's real work; nothing is ever sent automatically
• Honest about scale — public data only, through official APIs (GitHub, web search results) and no LinkedIn scraping. A run reads up to a few hundred people, and says so when GitHub reports more matches than it read
• Stop any time — everyone already judged is kept; past searches reopen as they ended and can be deleted with everyone they found

Hiring Manager Portal

• Resume Upload & RAG — PDF/DOCX parsing, ChromaDB vector storage, contextual AI chat
• Screening — scores every candidate against a role on fixed, inspectable weights and ranks them into a shortlist
• Resume Intelligence — Gap analysis, skill verification targets, red flag detection
• Scanner Agent — Extracts embedded links, scrapes GitHub profiles, searches LinkedIn via Tavily
• AI Chat — Multi-candidate comparison, voice input/output, anonymization mode
• Hiring Agent — Role-specific evaluation with JD matching, generates fit reports
• Credibility Analysis — Cross-references resume claims against interview performance
• Email Composer — AI-drafted emails (interest, interview, offer, pass, follow-up)
• Interview Creator — Configure role, level, questions, focus areas → grants candidate access
• PDF Report Export — Branded downloadable assessment reports

Candidate Portal

• Resume Upload — Single-resume-per-candidate with replace flow
• AI Advisor (3 Sub-Agents)
  • Resume Coach — Identifies gaps, suggests improvements, ATS optimization
  • Interview Prep — Practice questions, STAR method coaching, role-specific prep
  • Career Advisor — Strengths analysis, career paths, skill recommendations
• Live AI Interview — Real-time video interview with AI avatar
• Interview Report — Scores, eye contact %, proctoring summary, credibility analysis

Interview System

• LiveKit Cloud — WebRTC room infrastructure for real-time audio/video
• Simli Avatar — Lip-synced AI avatar as the interviewer's face
• OpenAI Realtime API — Voice-to-voice conversation (no TTS/STT latency)
• Smart Questions — Interview questions informed by resume gap analysis
• Proctoring — Face detection, eye tracking, tab monitoring, fullscreen enforcement
• 3-Violation Auto-Termination — Tab switch, window blur, or fullscreen exit = violation

Analytics

• Candidate Analytics — Skills distribution, experience comparison, role & level breakdown
• Interview Analytics — Completion rates, score distribution, high/low performers, recent interviews

UI/UX

• Glassmorphism Design — Backdrop blur, gradient borders, subtle animations
• Dark + Light Theme — Full support on both portals (amber accent hiring, blue accent candidate)
• DM Sans Typography — Consistent font system across all pages
• Responsive — Mobile-friendly sidebar collapse

Tech Stack

7 Specialized Agents

How It Works

Hiring Manager Flow

Manual (dashboard):

Upload Resumes → Data Agent parses + enriches
     ↓
Automate → AI ranks candidates, recommends interview order
     ↓
Resume Intel → Gap analysis, verification targets, red flags
     ↓
Hiring Agent → Evaluates against job description
     ↓
Create Interview → Smart questions from resume analysis
     ↓
Email Candidate → AI-drafted invitation
     ↓
Post-Interview → Credibility analysis + PDF export

Automated via Jarvis (voice or text):

"Screen everyone for a Python Developer"
     → Jarvis runs ATS, names top candidates
     ↓
"Check his GitHub" / "Give me his drawbacks"
     → Jarvis resolves the active candidate, reuses cached context when possible, and narrates findings
     ↓
"Show me her resume red flags" / "How did his interview go?"
     → Jarvis loads resume intelligence or the interview report directly in-chat
     ↓
"Set up interviews for the top 3"
     → Jarvis creates interviews, shows email draft card with Send button
     ↓
"Send it" / "Open the PDF report" → emails dispatch or export completes without leaving the conversation

Candidate Flow

Login (email verification) → access granted by hiring manager
     ↓
Upload Resume → AI analyzes and stores context
     ↓
AI Advisor → Resume Coach | Interview Prep | Career Advisor
     ↓
Join Interview → LiveKit room connects
     ↓
AI Avatar (Simli) interviews with resume-informed questions
     ↓
Report → scores, eye contact, credibility analysis, PDF export

License

All Rights Reserved. See LICENSE for details.

Built by Sai Punith Kolla