Hiring Outcome Predictor (AI Oracle & Reranker)

Grading API

The Grading API serves a two-sided large propensity model (LPM) that predicts a candidate's application intent and interview outcome for a given job. Trained on 1.2 billion hiring interviews and 400 million career paths, it surpasses human expert judgment (achieving 94% accuracy vs 64% for humans) to deliver the ultimate recruiting decision layer, with built-in fairness mitigation and EU AI Act explainability.

Read the Docs

94% accuracy

100ms/evaluation

43+ languages

GDPR & EU AI Act

99.99% uptime

output.json

{
  "grade": 0.94,
  "intent": "high",
  "outcome": "hired",
  "fairness": "compliant"
}

Trusted by Customers, Partners & the AI Ecosystem

API REQUEST

Query Params

Get fine-grained evaluation scores for application intent and interview success. Use it to rerank initial Scoring API results, maximize job description engagement, prioritize high-propensity hires, and eliminate low-yield interviews. Results are deterministic: same inputs → same scores.

Max batch size: 32 items/call (profiles or jobs).

Select Grading Type

Send a job + profiles → returns Profiles (JSON) + evaluation Scores.

Explore

API Guide API Endpoint Try Live Demo

Top Query Fields

Anchor Job

key or reference, and board

Grading Algorithm

Recruiter-side: Horus Cosmic, Horus Titan, Horus Turbo, Horus Nano

Profiles

list of profile IDs (key or reference, and source)

response.json

Returns: Profiles + Grading Scores

 1{
 2  "code": 200,
 3  "message": "Grading completed successfully.",
 4  "data": [
 5    {
 6      "profile": {
 7        "key": "abc123",
 8        "info": {
 9          "first_name": "Sarah",
10          "last_name": "Chen",
11          "location": "San Francisco, CA"
12        },
13        "skills": ["Python", "ML", "NLP"],
14        "experiences_duration": 7.2
15      },
16      "predictions": [0.06, 0.94]
17    },
18    {
19      "profile": {
20        "key": "def456",
21        "info": {
22          "first_name": "James",
23          "last_name": "Park",
24          "location": "New York, NY"
25        },
26        "skills": ["Data Science", "PyTorch"],
27        "experiences_duration": 5.1
28      },
29      "predictions": [0.13, 0.87]
30    }
31  ]
32}

🎯 MODEL SELECTOR

Grading Models by Use Case

Pick the right model version for your needs— by use case, speed, precision, and cost.

Algorithm key	Data	Use Case	Speed	Precision	Languages

Horus Cosmic	ProfileJob	Precision model optimized for deep contextual understanding of skills, experience, and cultural alignment. Best suited for accurate, high-stakes ranking in refined talent pools.	4th	1st	43 languages
Horus Titan	ProfileJob	Offers a robust equilibrium between speed and accuracy for reliable ranking performance across diverse candidate profiles.	3rd	2nd	43 languages
Horus Turbo	ProfileJob	Designed for rapid evaluation of large candidate volumes with solid predictive accuracy. Ideal for mass or program-based recruitment.	2nd	3rd	43 languages
Horus Nano	ProfileJob	Ultra-fast, lightweight model built to efficiently identify promising or interview-worthy candidates.	1st	4th	43 languages

CUSTOMER STORIES

Don't take our word for it!

Trusted by fast-growing HR Tech and Global Enterprise

“

No more candidate drop-off from long assessments

vs. HireVue

Before HrFlow.ai

Completion rates tanked — candidates were dropping off because the testing process was too long, and we were losing qualified people before recruiters could even evaluate them.

After HrFlow.ai

HrFlow.ai Grading solved that with ~94% accuracy and ~100ms response time, giving us deterministic recruiter-side and candidate-side scores in real time. We removed friction from the funnel and reduced abandonment.

“

No more prompt injection or keyword gaming

vs. OpenAI

Before HrFlow.ai

A clear ChatGPT writing bias meant resumes optimized with OpenAI scored higher regardless of quality. Prompt injection inside resumes became a real production risk.

After HrFlow.ai

HrFlow.ai Grading replaced that with a deterministic, HR-native model trained on hiring outcomes. We moved from manipulable keyword wins to stable, secure probability scores.

“

Reuse our talent pool, not retest it

vs. TestGorilla

Before HrFlow.ai

We couldn’t go back to thousands of passive candidates and ask them to retake long tests just to evaluate them for a new role.

After HrFlow.ai

HrFlow.ai Grading grades fit, application likelihood, and hiring likelihood directly from resumes and job descriptions—so we recycle our talent pool without adding friction.

“

No more candidate drop-off from long assessments

vs. HireVue

Before HrFlow.ai

Completion rates tanked — candidates were dropping off because the testing process was too long, and we were losing qualified people before recruiters could even evaluate them.

After HrFlow.ai

“

No more prompt injection or keyword gaming

vs. OpenAI

Before HrFlow.ai

A clear ChatGPT writing bias meant resumes optimized with OpenAI scored higher regardless of quality. Prompt injection inside resumes became a real production risk.

After HrFlow.ai

HrFlow.ai Grading replaced that with a deterministic, HR-native model trained on hiring outcomes. We moved from manipulable keyword wins to stable, secure probability scores.

“

Reuse our talent pool, not retest it

vs. TestGorilla

Before HrFlow.ai

We couldn’t go back to thousands of passive candidates and ask them to retake long tests just to evaluate them for a new role.

After HrFlow.ai

HrFlow.ai Grading grades fit, application likelihood, and hiring likelihood directly from resumes and job descriptions—so we recycle our talent pool without adding friction.

“

No more candidate drop-off from long assessments

vs. HireVue

Before HrFlow.ai

Completion rates tanked — candidates were dropping off because the testing process was too long, and we were losing qualified people before recruiters could even evaluate them.

After HrFlow.ai

“

No more prompt injection or keyword gaming

vs. OpenAI

Before HrFlow.ai

A clear ChatGPT writing bias meant resumes optimized with OpenAI scored higher regardless of quality. Prompt injection inside resumes became a real production risk.

After HrFlow.ai

HrFlow.ai Grading replaced that with a deterministic, HR-native model trained on hiring outcomes. We moved from manipulable keyword wins to stable, secure probability scores.

“

Reuse our talent pool, not retest it

vs. TestGorilla

Before HrFlow.ai

We couldn’t go back to thousands of passive candidates and ask them to retake long tests just to evaluate them for a new role.

After HrFlow.ai

HrFlow.ai Grading grades fit, application likelihood, and hiring likelihood directly from resumes and job descriptions—so we recycle our talent pool without adding friction.

“

No more candidate drop-off from long assessments

vs. HireVue

Before HrFlow.ai

Completion rates tanked — candidates were dropping off because the testing process was too long, and we were losing qualified people before recruiters could even evaluate them.

After HrFlow.ai

“

No more prompt injection or keyword gaming

vs. OpenAI

Before HrFlow.ai

A clear ChatGPT writing bias meant resumes optimized with OpenAI scored higher regardless of quality. Prompt injection inside resumes became a real production risk.

After HrFlow.ai

HrFlow.ai Grading replaced that with a deterministic, HR-native model trained on hiring outcomes. We moved from manipulable keyword wins to stable, secure probability scores.

“

Reuse our talent pool, not retest it

vs. TestGorilla

Before HrFlow.ai

We couldn’t go back to thousands of passive candidates and ask them to retake long tests just to evaluate them for a new role.

After HrFlow.ai

HrFlow.ai Grading grades fit, application likelihood, and hiring likelihood directly from resumes and job descriptions—so we recycle our talent pool without adding friction.

“

A more defensible grading layer

vs. Eightfold AI

Before HrFlow.ai

Legal issues around applicant scoring raised compliance concerns—opaque external profiling that functions like a credit-report-style score was not a direction we were comfortable taking.

After HrFlow.ai

HrFlow.ai relies only on candidate-submitted and recruiter-provided data. Deterministic scores, built-in explainability, fairness controls, and EU AI Act recognition.

“

Cosine reordering wasn’t hiring judgment

vs. Cohere Reranker

Before HrFlow.ai

Cohere reranked text relevance, not hiring outcomes. It lacked native concepts for interview success, application intent, and recruiter-side qualification.

After HrFlow.ai

HrFlow.ai Grading gave us deterministic, HR-native probability grading trained on hiring signals—the difference between ‘this looks relevant’ and ‘this person is most likely to succeed.’

“

We outgrew taxonomies and keyword overlap

vs. AdeptID

Before HrFlow.ai

Elasticsearch gave us endless manual tuning but never true hiring likelihood. We kept patching the system with taxonomies, filters, synonyms, and edge-case rules.

After HrFlow.ai

HrFlow.ai Grading replaced that brittle stack with an HR-native model that reasons over full profile and job context—deterministic probability scores with better precision on complex roles.

“

A more defensible grading layer

vs. Eightfold AI

Before HrFlow.ai

Legal issues around applicant scoring raised compliance concerns—opaque external profiling that functions like a credit-report-style score was not a direction we were comfortable taking.

After HrFlow.ai

HrFlow.ai relies only on candidate-submitted and recruiter-provided data. Deterministic scores, built-in explainability, fairness controls, and EU AI Act recognition.

“

Cosine reordering wasn’t hiring judgment

vs. Cohere Reranker

Before HrFlow.ai

Cohere reranked text relevance, not hiring outcomes. It lacked native concepts for interview success, application intent, and recruiter-side qualification.

After HrFlow.ai

HrFlow.ai Grading gave us deterministic, HR-native probability grading trained on hiring signals—the difference between ‘this looks relevant’ and ‘this person is most likely to succeed.’

“

We outgrew taxonomies and keyword overlap

vs. AdeptID

Before HrFlow.ai

Elasticsearch gave us endless manual tuning but never true hiring likelihood. We kept patching the system with taxonomies, filters, synonyms, and edge-case rules.

After HrFlow.ai

HrFlow.ai Grading replaced that brittle stack with an HR-native model that reasons over full profile and job context—deterministic probability scores with better precision on complex roles.

“

A more defensible grading layer

vs. Eightfold AI

Before HrFlow.ai

Legal issues around applicant scoring raised compliance concerns—opaque external profiling that functions like a credit-report-style score was not a direction we were comfortable taking.

After HrFlow.ai

HrFlow.ai relies only on candidate-submitted and recruiter-provided data. Deterministic scores, built-in explainability, fairness controls, and EU AI Act recognition.

“

Cosine reordering wasn’t hiring judgment

vs. Cohere Reranker

Before HrFlow.ai

Cohere reranked text relevance, not hiring outcomes. It lacked native concepts for interview success, application intent, and recruiter-side qualification.

After HrFlow.ai

HrFlow.ai Grading gave us deterministic, HR-native probability grading trained on hiring signals—the difference between ‘this looks relevant’ and ‘this person is most likely to succeed.’

“

We outgrew taxonomies and keyword overlap

vs. AdeptID

Before HrFlow.ai

Elasticsearch gave us endless manual tuning but never true hiring likelihood. We kept patching the system with taxonomies, filters, synonyms, and edge-case rules.

After HrFlow.ai

HrFlow.ai Grading replaced that brittle stack with an HR-native model that reasons over full profile and job context—deterministic probability scores with better precision on complex roles.

“

A more defensible grading layer

vs. Eightfold AI

Before HrFlow.ai

Legal issues around applicant scoring raised compliance concerns—opaque external profiling that functions like a credit-report-style score was not a direction we were comfortable taking.

After HrFlow.ai

HrFlow.ai relies only on candidate-submitted and recruiter-provided data. Deterministic scores, built-in explainability, fairness controls, and EU AI Act recognition.

“

Cosine reordering wasn’t hiring judgment

vs. Cohere Reranker

Before HrFlow.ai

Cohere reranked text relevance, not hiring outcomes. It lacked native concepts for interview success, application intent, and recruiter-side qualification.

After HrFlow.ai

HrFlow.ai Grading gave us deterministic, HR-native probability grading trained on hiring signals—the difference between ‘this looks relevant’ and ‘this person is most likely to succeed.’

“

We outgrew taxonomies and keyword overlap

vs. AdeptID

Before HrFlow.ai

Elasticsearch gave us endless manual tuning but never true hiring likelihood. We kept patching the system with taxonomies, filters, synonyms, and edge-case rules.

After HrFlow.ai

HrFlow.ai Grading replaced that brittle stack with an HR-native model that reasons over full profile and job context—deterministic probability scores with better precision on complex roles.

🔗 INTEGRATIONS

Works with the tools you use

Integrate 200+ tools with the flip of a switch.

HrFlow.ai Data Studio

HR-native ETL with 200+ connectors plus Webhooks to ingest, normalise, and sync jobs & profiles across your stack—reliable pipelines with unified schemas.

Zapier ETL

No-code automation platform with 8,000+ app integrations to move data between tools using triggers + actions.

Make.com ETL

Visual automation platform to extract/transform/route data across 3,000+ apps (plus HTTP modules for any API).

Microsoft Flow ETL

Microsoft Power Automate—workflow automation with 1,000+ API connectors (and support for custom connectors).

Workato ETL

Enterprise iPaaS/automation platform with 1,200+ pre-built connectors for orchestrating integrations and data workflows at scale.

Salesforce Flow Automation

Salesforce's low-code workflow automation tool; extended via AppExchange with 7,000+ apps to add integrations and capabilities.

🚀 KEY FEATURES

End-to-End Grading Stack: High-Precision Assessment, Explainability, and Responsible AI

HrFlow.ai Grading is a large propensity model (LPM) for People ↔ Jobs evaluation, built for maximum precision and reaching ~94% accuracy on hard HR decision tasks. It uses HR-native deep hierarchical graders trained on full career trajectories, job requirements, context, work environment, and hiring signals. The API delivers deterministic probability scoring with multilingual and cross-lingual grading across 43+ languages. It also supports role and industry generalization, pretrained graders selectable by speed versus precision, explainability (Reasoning API), and feature fusion using tags, metadata, scorecards, behavioral signals, and tabular data. The system is powered by specialized candidate-side and recruiter-side graders trained with built-in fairness regularization and representation-bias calibration, aligned with the EU AI Act and international ethical AI requirements.

Tutorial Video3:45

🔒 ENTERPRISE-READY

Trust & Security

Built for sensitive HR data—secure by default, enterprise-ready.

Encryption

TLS in transit + encryption at rest to protect documents and extracted data.

Retention control

Minimal storage by default, with configurable retention policies to match your compliance needs.

AI-Act / GDPR / DPA ready

Built for sensitive HR data—secure by default, enterprise-ready. AI Act– and GDPR-ready processing, with documented controls for data handling and compliance.

Location / Region

Data processing and storage can be aligned with your required region (e.g., EU or US) depending on your deployment.

Visit Trust Center→

📊 FEATURE COMPARISON

HrFlow.ai Grading is the fastest, most accurate, and only outcome-based assessment API for People & Jobs

Feature	HrFlow.ai Grading	HireVue	OpenAI LLM	TestGorilla	Eightfold AI	Harver	AdeptID	Cohere reranker	ElasticSearch

Deployment & Trust
Headquarters	🇫🇷 France	🇺🇸 USA	🇺🇸 USA	🇳🇱 Netherlands	🇺🇸 USA	🇳🇱 Netherlands	🇺🇸 USA	🇺🇸 USA	🇺🇸 USA
🇺🇸 USA & 🇪🇺 EU Servers	Built-in	config	config	Built-in	config	Built-in			config
GDPR / AI-Act readiness	By design			By design		By design			config
HR Compliance (Safety & Guardrails)	Built-in	(Class Action)		Built-in	(Class Action)	Built-in	Limited
Pretraining Data	1,2B Hiring Signals (Top Hiring firms)	No outcome-based training	Noisy & Biased Web Data	No outcome-based training	Noisy Corporate Data & Social Profiling	No outcome-based training	No outcome-based training	Noisy & Biased Web Data	No outcome-based training
HR-Focused
Input Security (Prompt injection)
Unified output object (JSON drift)
Deterministic output				(Mood influence)		(Mood influence)
Pricing model	per request	per subscription (bundle)	per input+output tokens (expensive)	per test	per subscription (bundle)	per subscription (bundle)	per record	per input tokens + record (expensive)	per Server (unpredictable)
Speed (avg response / 1 application)	~100ms	~20min	~5s	~60min	~2min	~35min	~2s	~300ms	~1s
Scalability	Real-time API	Requires candidates	Real-time API	Requires candidates	Requires candidates	Requires candidates	Real-time API	Real-time API	Real-time API
UI Access	Built-in	Built-in		Built-in	Built-in	Built-in	Built-in
Core Technology
Technology	Large Propensity Model / Fairness & Bias Optimization	Skills tess / game-based psychometrics, AI-scored video interviews, virtual job tryouts	LLM	Skills tests, cognitive tests, personality /psychometric tests	LLM / Deep Encoders	Skills tests, cognitive tests, personality /psychometric tests	Custom ElasticSearch	Deep Encoders	Apache Lucene / TF-IDF
Multilingual	43 lang	40 lang	40 lang	12 lang	24 lang	12 lang	EN pivot	23 lang	Config
Crosslingual
Match Scores	Success Probability	Criteria Overlap	Concepts Overlap	Criteria Overlap	Criteria Overlap	Criteria Overlap	Keywords Overlap	Cosine Similarity	Keywords Overlap
Two-side Grading	Asymmetric Score (Candidate-side+ Recruiter-side)	Recruiter Score Only	Config	Recruiter Score Only	Symetric Score (Candidate-side = Recruiter-side)	Recruiter Score Only	Symetric Score (Candidate-side = Recruiter-side)	Config	Config
Hiring Likelihood	Built-in (Outcome-based grading)	No Feedback loop	Not trainable	No Feedback loop	Built-in (Outcome-based grading)	No Feedback loop	Not trainable	Needs Signal Data	Not trainable
Application Likelihood	Built-in (Outcome-based grading)	Not trainable	Not trainable	Not trainable	Built-in (Outcome-based grading)	No Feedback loop	Not trainable	Needs Signal Data	Not trainable
White-collar Roles Accuracy	High	Good	Good	Good	Medium	Good	Low	Medium	Low
Blue-collar Roles Accuracy	High	Good	Medium	Medium	Low	Medium	Low	Low	Low
Junior Roles Accuracy	High	Medium	Medium	Medium	Low	Medium	Low	Low	Low
Senior Roles Accuracy	High	Medium	Good	Medium	Medium	Medium	Low	Low	Low
Custom Feature Engineering	Built-in (Tags & Metadata)	Built-in (Custom questions)	Concept boosting	Built-in (Custom questions)	Built-in (Custom criteria)	Built-in (Custom questions)	Built-in (Custom criteria)		Keyword boosting
Fairness Regularization	Built-in (Constraints)
Data Calibration & Debiasing	Built-in (Pipeline)
Performance Monitoring	Built-in (HR-native)
Fairness & Bias Monitoring	Built-in (HR-native)
HR Stack integrations (add-ons)
Hybrid Search	Built-in (Searching API)	Built-in		Built-in	Built-in	Built-in	Built-in	Built-in	Config
Reasoning & Explainability	Built-in (Reasoning API)		Built-in (matched concepts)						Built-in (matched keywords)
Resume, CV, Job parsers	Built-in (Parsing API)	Third party	Config	Third party	Third party	Third party	Third party
HR data enrichment & taxonomies	Built-in (Linking/Tagging/Asking APIs)				Built-in		Limited
Browser Extension	Connector (Data Studio)				Built-in
Jobboards / ATS / HCM / HRIS connectors	200+ connectors (Data Studio)	Built-in		Built-in	Built-in	Built-in	Limited
Candidate & Recruiter UI	Widgets (App Studio)	Built-in		Built-in	Built-in	Built-in

❓ COMMON QUESTIONS

Frequently Asked Questions

Everything you need to know about the Grading API

🧩 COMPLETE API SUITE

Go beyond the Grading API

Our APIs are designed to complement each other and unlock your data's full potential

Full Extraction API Suite

Transform HR documents into structured, enriched Talent & Workforce Data — powering every layer of Hiring Intelligence.

API Overview

Ranking API Suite

Unlock Hiring Superintelligence at scale — with transparent, fair, and explainable ranking across every Talent signal.

API Overview

Grading API

Query Params

Select Grading Type

Explore

Top Query Fields

Grading Models by Use Case

Don't take our word for it!

No more candidate drop-off from long assessments

No more prompt injection or keyword gaming

Reuse our talent pool, not retest it

No more candidate drop-off from long assessments

No more prompt injection or keyword gaming

Reuse our talent pool, not retest it

No more candidate drop-off from long assessments

No more prompt injection or keyword gaming

Reuse our talent pool, not retest it

No more candidate drop-off from long assessments

No more prompt injection or keyword gaming

Reuse our talent pool, not retest it

A more defensible grading layer

Cosine reordering wasn’t hiring judgment

We outgrew taxonomies and keyword overlap

A more defensible grading layer

Cosine reordering wasn’t hiring judgment

We outgrew taxonomies and keyword overlap

A more defensible grading layer

Cosine reordering wasn’t hiring judgment

We outgrew taxonomies and keyword overlap

A more defensible grading layer

Cosine reordering wasn’t hiring judgment

We outgrew taxonomies and keyword overlap

Works with the tools you use

HrFlow.ai Data Studio

Zapier ETL

Make.com ETL

Microsoft Flow ETL

Workato ETL

Salesforce Flow Automation

End-to-End Grading Stack: High-Precision Assessment, Explainability, and Responsible AI

Probability match scores

Jobs Grading (Application Likelihood)

Profiles Grading (Hiring Likelihood)

43+ Languages (Multilingual)

Cross-Lingual Grading

Role & Industry Generalization

Pretrained Graders Marketplace

Custom Feature Engineering

Reranker for Scoring API (The Winning Duo)

Reasoning API (Explainability)

Fairness-Regularized Encoders

Representation-Bias Calibrated Encoders

Trust & Security

Encryption

Retention control

AI-Act / GDPR / DPA ready

Location / Region

HrFlow.ai Grading is the fastest, most accurate, and only outcome-based assessment API for People & Jobs

Frequently Asked Questions

What is the Grading API?

Where can I use the Grading API?

What does "large propensity model (LPM)" mean?

What does "two-sided grading" mean?

How is Grading different from Searching, Matching, and Scoring APIs?

What is a reranker?

How many languages do you support?

What's the difference between multilingual and cross-lingual?

How do I grade Profiles against a Job?

How do I grade Jobs for a candidate?

Are results deterministic?

What is Hiring Likelihood?

Can the same Profile–Job pair receive different scores?

What level of accuracy does the Grading API reach?

Can the Grading API work across industries and job types?

What kind of training data supports the Grading API?

How do I choose between Cosmic, Titan, Turbo, and Nano?

How do I build a custom model with my own data (e.g., salary, scorecards)?

How do I build a custom model with Signals?

How do I measure grading accuracy on both sides (Profiles ↔ Jobs)?

How do I combine Scoring and Grading into a single workflow?

How do I explain scores to recruiters and candidates?