Knowledge graphs that outperform standard LLMs and naive RAG.

Naive RAG chops textbooks into arbitrary text chunks, scoring 72.1 on complex STEM. Hesperus connects curricula into a computable prerequisite DAG—boosting benchmark performance to 84.1 with 100% hallucination resistance and 28% fewer tokens.

Try the Live App
Illustrative Examples · Retrieval and Prerequisite Context
Inquiry (Physics • Mechanics):

“Why does static friction do zero net work in pure rolling without slipping?”

Naive Vector RAG + Standard LLM

Cosine similarity over unlinked text chunks

Naive vector RAG injects 1,800+ tokens of fragmented text. The LLM guesses that friction always dissipates energy, confusing contact point constraints with kinetic sliding.

Benchmark outcome: 72.14 composite score; severing equations and wasting reasoning tokens on noisy text chunks.
Hesperus Knowledge Graph Engine
Prerequisite Graph Traversal:
Instantaneous axis of rotation & zero relative velocity at contact
Axiomatic Resolution:“At the exact instant of contact during pure rolling, what is the relative velocity of the contact point? If instantaneous displacement during contact is zero, what is the work done?”
Suggested learning connection:Proves net work is strictly zero via the work-energy theorem (Wtrans=−fsdW_{\text{trans}} = -f_s d, Wrot=+fsdW_{\text{rot}} = +f_s d).

Knowledge graphs structured for rigorous STEM disciplines and production platforms

Mechanics & Physics
Rotational dynamics, electrostatics, thermodynamics DAG
Calculus & Algebra
Multivariable integrals, differential equations, vectors
Chemistry Topology
Kinetics, molecular orbitals, reaction mechanisms
High-Stakes Exams
JEE Advanced, NEET, Olympiads, and AP/IB STEM
Live Platform
Powering queries live at app.hesperusai.com
Graph TraversalSub-millisecond prerequisite path resolution
Data IsolationDedicated private graph enclaves per client
Formula GroundingSource formulas & linked diagrams
Production SpeedLive on your custom domain in under 48 hours
Why Graph Beats RAG

Why naive RAG fails on
complex STEM knowledge.

Textbooks and curricula are not bags of text snippets. They are hierarchical dependency trees where advanced reasoning requires verified foundational axioms.

The Flaw of Naive Vector RAG

Vector similarity is blind to prerequisite hierarchy.

Standard RAG slices documents into arbitrary chunks and retrieves text by cosine similarity. In complex STEM, this fails: in our 90 controlled trials, Chunk RAG scored 72.14—worse than raw baseline models (73.39).

Injecting ~1,800 tokens of fragmented text severs mathematical derivations mid-equation and forces reasoning models to waste their internal Chain-of-Thought budget denoising text rather than solving problems.

Empirical failure modes of naive RAG:
• Drops composite benchmark score to 72.14 (below unaugmented base LLM)
• Injects 1,800+ tokens of fragmented text that cuts equations mid-proof
• Wastes reasoning tokens filtering noise rather than deriving solutions
The Hesperus Knowledge Graph

Deterministic concept topology and traversal.

Hesperus converts curricula into a Directed Acyclic Graph (DAG) of canonical concepts, verified LaTeX formulas, and prerequisite edges. When queried, our engine traverses the graph upstream in 0.27ms to retrieve exact foundational axioms.

By delivering deduplicated, high-density concept containers rather than noisy text chunks, models achieve a 28% prompt token reduction and spend 100% of their reasoning budget on rigorous derivation.

Empirical benchmark advantages:
• Leaps composite score from 71.8 to 84.1 on lightweight $0.10/M models
• 100% hallucination resistance across canonical STEM problem benchmarks
• +40% real exam grounding with 28% prompt token reduction

We replace fragile vector chunking with computable knowledge graphs. Experience the difference live at app.hesperusai.com.

Platform Architecture

Topological concept graphs,
not stochastic vector guessing.

Three core architectural layers that make Hesperus outperform naive vector RAG and generic LLMs on complex STEM reasoning.

DAG Engine
GRAPH RETRIEVAL

Topological Prerequisite DAG

Naive RAG computes cosine similarity between text chunks, failing when a student has an unstated prerequisite deficit. Hesperus traverses a directed acyclic graph upstream to isolate the root conceptual gap before generating guidance.

Core advantages:
Recursively isolates upstream conceptual deficits
0.27ms in-memory prerequisite graph traversal
28% prompt token reduction vs naive chunk RAG
Eliminates reasoning token waste on noisy text parsing
Axiomatic Grounding
ACCURACY & CITATIONS

Axiomatic Grounding & Socratic Hints

Standard LLMs hallucinate formulas and dump copy-paste solutions. Hesperus binds every concept node to verified LaTeX axioms, coordinate geometry, and Socratic hint ladders that nudge students forward without giving away the answer.

Core advantages:
100% hallucination resistance in 90 controlled trials
+40% real exam grounding via direct PYQ linkage
LaTeX axioms and formula inheritance from upstream nodes
Adaptive reasoning budgets ensuring complete derivations
Private Enclave
DATA SOVEREIGNTY

Private Tenant Graph Enclaves

Your question banks, mock exams, and lecture notes are retrieved from the configured institute workspace. Relevant excerpts are sent to inference providers with routing that requests no data collection; review provider terms before sharing proprietary material.

Core advantages:
Dedicated graph database partition per client
Provider routing requests no data collection; provider terms apply
Isolated REST API endpoints with low-latency SLAs
Enterprise VPC and on-premise deployment options
Live Reference Implementation

Experience the Knowledge Graph live in production

Interact directly with our live reference platform at app.hesperusai.com. Test physics, calculus, and chemistry problems across our real curriculum graph.

Launch Live App
84.1
Benchmark Score
Outperforms raw $20/M frontier models with lightweight $0.10/M models
100%
Hallucination Resistance
Zero invented formulas or spurious constants across 90 controlled trials
+40%
Real Exam Grounding
Direct linkage to past competitive exam problem patterns
-28%
Token Footprint
Canonical DAG containers vs bloated, fragmented Chunk RAG
Benchmark Verification · 90 Controlled Trials

Composite score evaluation against official IIT-JEE Advanced challenge sets.

Normal LLM (Baseline)
AxiomGraph (Knowledge Graph)+11.4% Avg
606570758085909571.884.1(+12.3)Gemini 2.5 FlashLightweight Frontier73.779.2(+5.5)gpt-5.4-miniMedium Frontier74.782.1(+7.4)gpt-5.4High Frontier73.481.8(+8.4)Global Benchmark AverageAll 90 Controlled Trials
Google Gemini 2.5 FlashLightweight Frontier

AxiomGraph empowers sub-$0.10/M models to beat raw $20/M frontier models.

Score Delta
+12.3 (+17.1%)
Controlled Trials
30 tests
Hallucination
0.0%
Token Reduction
-31%
Integration Roadmap

From live evaluation
to deep graph API integration.

Test the engine immediately on app.hesperusai.com, deploy a 48-hour pilot on your subdomain, or integrate directly with our low-latency graph APIs.

PHASE 01
Days 1–30 (48-hour setup)

Live Sandbox & Evaluation

app.hesperusai.com or custom domain

Evaluate the Knowledge Graph immediately on app.hesperusai.com or deploy a branded private sandbox on your custom subdomain (e.g. graph.yourinstitute.com). Faculty and academic heads test prerequisite navigation on real curricula.

Evaluate live platform at app.hesperusai.com
Custom subdomain sandbox live in < 48 hours
Ingest sample syllabus DAG or question bank
Zero client app code modification required
Rollout stageStage 1 of 3
PHASE 02
Month 2

REST API & In-App Graph Room

Embed via SDK or JWT WebView

Connect your mobile app or LMS to our low-latency REST graph API or drop in an authenticated learning room inside your React Native, Flutter, iOS, or Android app with seamless single sign-on.

Sub-millisecond REST API for DAG retrieval
Drop-in iframe / WebView with SSO JWT tokens
Real-time student concept deficiency telemetry
Faculty dashboard with prerequisite drop-off analytics
Rollout stageStage 2 of 3
PHASE 03
Scale Phase

Native SDK & Automated Test Linking

Convert test mistakes to remedial paths

Deeply link your mock test engine with our Knowledge Graph. When students miss an exam question, automated webhooks query the DAG to assign targeted prerequisite practice sets.

Automated exam error to prerequisite DAG linking
Camera-based math & diagram input parser
Cohort-level concept mastery heatmaps
Dedicated VPC or on-premise graph hosting
Rollout stageStage 3 of 3
Enterprise Licensing & Collaboration

Predictable Enterprise Economics.
Engineered for Global Scale.

Transparent per-learner licensing with zero token surprises. Deploy a private enterprise sandbox in 48 hours.

Research Pilot Sandbox

Zero-Risk Evaluation

For institutions and platforms evaluating neuro-symbolic reasoning on a single trial cohort.

₹014-day dedicated sandbox
Architecture Deliverables:
Up to 200 active learners (1 cohort)
Branded subdomain (ai.yourplatform.com)
1-Click Faculty DPP generator testbed
Full Physics, Math & Chemistry DAG access
Zero code modifications to existing mobile apps
Architecture onboarding from IITD
Featured Platform Tier

Enterprise Platform

Mentor API

Production infrastructure layer for hybrid coaching networks and global EdTech apps.

₹40per learner / month
Architecture Deliverables:
Everything in Research Pilot Sandbox
Full Hesperus Mentor API access (REST & Webhooks)
In-App WebView Tab via cryptographically signed JWT
Complete institutional question bank ingestion
Institute-scoped curriculum access
Branded practice-sheet drafts with browser PDF export
Real-time cohort prerequisite drop-off heatmaps
99.9% uptime SLA with multi-provider failover

Sovereign Infrastructure

Private VPC / On-Prem

For national educational boards, universities, and high-security enterprise networks.

Custominstitutional SLA agreement
Architecture Deliverables:
Everything in Enterprise Platform
Dedicated isolated VPC or On-Premise vLLM cluster
Custom curriculum prerequisite DAG engineering
Native camera equation OCR SDK endpoints
Bi-directional ERP / LMS telemetry webhooks
Dedicated solutions architect (IITD)
Custom Socratic hint latency thresholds
Enterprise 99.99% high-availability SLA
FAQ

Frequently Asked Questions

Straightforward technical and commercial answers about our Knowledge Graph engine, accuracy limitations, and integration.