MambaHR Research

We don't wrap models.
We build them.

MambaHR Research is a domain AI lab focused on HR decision automation and people operations. We publish benchmarks, build proprietary datasets, and file patents — because HR deserves more than a ChatGPT wrapper.

HRIS dataencodeagent coreverifyaction
95K
Lines of code
500+
Benchmark scenarios
7
Patents pending

Research areas

01

Policy Verification

Deterministic statutory logic that verifies every HR decision against encoded HR policies and regulations. 95K lines of domain rules — not a language model guess, but a verified, auditable answer.

Statutory LogicPolicy EngineRegulation MappingAudit Trail
02

HR Decision Automation

Autonomous agent architecture that handles complete HR workflows end-to-end — from intake to resolution — without human intervention. Designed for safety-critical people operations.

Agent ArchitectureWorkflow OrchestrationHRIS Integration
03

Workforce Intelligence

Predictive models for attrition risk, compensation gap analysis, and headcount planning. Surfacing insights that HR teams would otherwise discover too late.

Attrition ModelingComp AnalysisHeadcount Forecasting

Publications

Our work

In preparationTarget: arXiv + EMNLP workshop· Q2 2026

HR-Bench: Evaluating AI on HR Decision-Making

500 scenarios across federal and state HR regulations.

In preparationTarget: ICLR workshop· Q3 2026

Deterministic Verification Layers for Agentic AI in HR

In preparationTarget: FAccT· Q4 2026

Calibrated Uncertainty Quantification for HR AI

HR-Bench

The first open benchmark for AI on HR decision-making.

500 scenarios · Federal + state regulations · Open dataset

7 Patents Pending

Novel research in HR AI that doesn't exist anywhere else.

01Decision confidence quantification for HR automation
02Workforce delta intelligence and anomaly detection
03HR decision archaeology and audit trail reconstruction
04Multi-jurisdictional policy conflict resolution
05Autonomous onboarding workflow orchestration
06Real-time policy update propagation without catastrophic forgetting
07HR agent confidence calibration with human escalation thresholds

Join the lab.

We're looking for ML engineers, NLP researchers, and HR domain experts who want to build AI that actually works in the real world.

View open roles