// EVALUATION.API.LAYER
Evaluation APIs That Return Learning Signals, Not Just Scores
CrazyGoldFish Evaluation APIs expose an infrastructure layer that captures learning signals during evaluation, structures them into usable data, and stores them as persistent learning memory.
POST /v1/evaluate → returns LearningSignal[]// FIELD.VALIDATION
// SIGNAL.FLOW
Every evaluated response becomes structured learning memory.
The Evaluation APIs convert evaluation into structured learning signals that systems can use.
01
Capture learning signals
- • Question-level evaluation across handwritten, typed, and multimodal responses
- • Concept understanding, step-level reasoning, rubric adherence
- • Extraction of intent, correctness, and reasoning patterns
02
Structure signals
- • Standardized, machine-readable outputs
- • Question-level granularity with rich metadata
- • Comparable across students, exams, and cohorts
03
Store as memory
- • Persistent learning records
- • Reusable across workflows and systems
- • Builds longitudinal learning intelligence
EVALUATION WORKFLOW → SIGNAL LAYER → STORAGE LAYER → MEMORY LAYER
// HOW IT WORKS
Submit a response.
Get back a learning signal.
Submit
Upload handwritten, typed, or multimodal responses via API
Evaluate
Apply rubric-based evaluation with policy configuration
Structure
Receive machine-readable signals at question-level granularity
Store
Persist learning memory — reusable across workflows and systems
Evaluation becomes a programmable layer inside your system.
SUBMIT
Response received via API
EVALUATE
Rubric-based evaluation applied
STRUCTURE
Signal extracted and formatted
STORE
Persisted as learning memory
// signal output
signal_type: concept_gap
confidence: 0.94
memory_id: stud_7f2a_032
// GOVERNED.DEPLOYMENT
Built for controlled, reliable evaluation
This is not blind automation. The Evaluation APIs are designed for governed deployment.
Human-in-the-loop
Teachers or reviewers remain final authority. Overrides and edits are fully supported.
Full audit trails
Every evaluation decision is logged. Full question-level traceability.
Adjudication & overrides
Disagreements are surfaced, not hidden. Structured resolution workflows.
Deterministic flagging
Outliers, low-confidence responses, and hotspots routed for review automatically.
"Evaluation can scale without losing control."
// BUSINESS IMPACT
Turning evaluation into infrastructure changes how learning systems operate.
Consistency
Standardised scoring across evaluators and cohorts
Auditability
Full traceability from response → signal → decision
Scalability
Evaluate large volumes without proportional increase in manual effort
Personalization
Structured learning signals power remediation and adaptive learning
Without structured signals, learning systems cannot improve.
// WHY NOW
Evaluation has become computable
Multimodal AI can interpret complex student responses
Subjective answers can now be evaluated reliably
Question-level signal capture is possible at scale
The constraint was not data — it was interpretation. That constraint is now removed.
This enables evaluation to function as infrastructure.
// BUILT.FOR.YOUR.CONTEXT
Which type of organization are you?
CrazyGoldFish is built for organizations that deliver learning at scale. Find your implementation context.
EdTech Platforms
Adaptive learning products that need evaluation signals at product scale.
See how it works →Coaching & Test-Prep Networks
High-stakes exam preparation where scoring precision determines outcomes.
See how it works →LMS / ERP Platforms
Infrastructure providers embedding evaluation into existing institutional workflows.
See how it works →Institutions & Public Programs
Government programs and NGOs delivering learning outcomes at district or national scale.
See how it works →// COMMON.QUESTIONS
CrazyGoldFish Evaluation APIs expose evaluation infrastructure via REST APIs. Teams integrate directly into LMS, ERP, or custom platforms to capture learning signals, structure them into machine-readable outputs, and store persistent learning memory at question-level granularity.
What do the Evaluation APIs return?
The APIs return structured learning signals at question-level granularity — including concept understanding, step-level reasoning, rubric adherence, confidence scores, and extracted intent. Outputs are machine-readable JSON, comparable across students, exams, and cohorts.
Can the APIs evaluate handwritten responses?
Yes. The Evaluation APIs support handwritten, typed, and multimodal responses. Multimodal AI interprets complex answer formats, including diagrams and mixed-media submissions.
How does human-in-the-loop work with the APIs?
Every evaluation decision can be routed for human review. Teachers or reviewers remain the final authority. The APIs surface disagreements, support overrides and edits, and log all decisions for full auditability.
What accuracy can we expect from the evaluation?
93% reliability in total scoring alignment (ICC). AI adjudicated correctness is 89.7% vs teacher 82.8% — a +6.9% improvement. 87% of responses fall within tolerance (±6 marks). Validated at India AI Impact Summit 2026, MeitY.
How do the Evaluation APIs integrate into an existing LMS or ERP?
Integration follows a 4-step flow: submit responses via API, trigger rubric-based evaluation with policy configuration, receive structured signal outputs, then store or act on signals within your existing system. The APIs are designed to embed directly into existing learning infrastructure — not replace it.
// RELATED · RESEARCH & PLAYBOOKS
// START.YOUR.DEPLOYMENT
Build your evaluation infrastructure
Integrate Evaluation APIs to capture learning signals at scale.