NexHR
Multi-model AI in a single HR platform
NexHR is an AI-powered HR management system that puts four distinct ML systems — semantic resume screening, biometric attendance, an automated payroll engine, and a RAG HR assistant — into one integrated, production-grade product with five roles and full RBAC.
ContextFinal Year Project — BS Computer Science, COMSATS University Islamabad, Lahore Campus (2026). Built to production-grade standards; demonstrates Bitlogs' AI integration capability.
4
AI/ML integrations
5
RBAC user roles
0
Manual payroll steps
Inside the product.
A walkthrough of the actual interfaces — merchant portal, operations console, and the two mobile apps shipped end-to-end.
Dashboard Overview
What was broken, risky, or missing.
Managing employees manually — tracking attendance, processing payroll, screening candidates — is time-consuming, error-prone, and doesn't scale.
Existing platforms (Zoho, Workday, Gusto) automate the basics but don't offer AI-driven screening, biometric attendance, and an intelligent HR assistant in one system.
The decisions — and the trade-off.
The goal was to prove that multiple AI subsystems can live inside one coherent product — not four disconnected tools bolted together.
Semantic resume screening using BAAI/bge-base-en-v1.5 (768-dim embeddings) with cosine-similarity ranking.
Biometric attendance via an OpenCV + ONNX Runtime face-recognition pipeline with geofencing and a manual fallback.
A service-aggregation payroll engine (attendance + loans + expenses) with gross-to-net tax brackets and Stripe Connect disbursements.
A RAG HR chatbot on Google Gemini 2.0 Flash, indexing company policy documents for 24/7 self-service.
Five distinct RBAC roles (Admin · HR Manager · Finance Manager · Employee · Candidate) behind JWT auth.
The Trade-off
Integrating four AI subsystems into one product is far harder than shipping any one of them alone — but a unified system is the entire point, and the only honest demonstration of production AI integration.
Concrete, shipped, documented.
Six modules, four AI/ML integrations, one product — fully tested.
AI / ML Modules
- AI Recruitment — semantic screening, automated interview scheduling, LinkedIn API integration
- Biometric Attendance — OpenCV + ONNX face recognition, cosine matching, geofencing, manual fallback
- RAG HR Chatbot — Gemini 2.0 Flash over indexed policy docs, 24/7 employee self-service
Platform
- Payroll Engine — service aggregation, gross-to-net tax, payslip snapshots, Stripe Connect
- Role-Based Dashboards — 5 roles with JWT authentication
- Leave & Financial Reporting — leave workflow, PDF payslips, audit-ready finance reports
AI Scoring Formula
0.50 · 0.30 · 0.20
Final Score = (0.50 × Semantic Similarity) + (0.30 × Skills Coverage) + (0.20 × Experience Match). Semantic similarity via cosine distance between BAAI embeddings of the job description and candidate profile.
Before and after — honestly.
NexHR is proof that Bitlogs can design and deploy multi-model AI systems in a single, integrated product. The same engineering decisions — ONNX optimization, RAG pipelines, pgvector semantic search, async task queues — apply directly to client AI products. This is not tutorial-level AI; it's production architecture.
The same engineering decisions behind NexHR — ONNX optimization, RAG pipelines, pgvector semantic search, async task queues — apply directly to client AI products. This is production architecture, not a tutorial.
Want results like NexHR?
Tell us what you're building. We'll tell you how long it takes and what it costs — for free, in plain English.
No agency jargon. No surprise invoices. Just engineers who give you a straight answer.