Rubber Supply Chain Multilingual Document Extraction with Claude Agents & Sarvam AI
Malaysia is advancing innovation in its rubber industry, digitizing multi-regional supply chain documentation across Southeast Asia and South Asia. Raw rubber procurement documentation spans multiple languages (Malay, English, Tamil, and Hindi), physical receipts, and trade certificates. In this challenge, you will implement an end-to-end supply chain inspection agent using Claude Agents SDK integrated with Sarvam AI's specialized Indic language models. The agent receives multimodal receipts, translates regional terminology, extracts technical rubber grade metrics (e.g., SMR 20 viscosity, moisture Content), and flags compliance breaches against international export standards.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a multilingual rubber supply chain extraction agent combining Claude Agents SDK with Sarvam AI Indic language processing.
How work is evaluated
Evaluates accurate multilingual field extraction and technical compliance verification.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
extraction_precision
Extracted specifications must match standard schema.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
translation_bleu_score
Quality score for domain-specific technical terminology translation • target: 0.88 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Construct Claude Agents SDK client workflows using extended thinking for complex reasoning
Integrate Sarvam AI APIs for regional Indic and Southeast Asian language translation
Extract structured technical rubber properties (DRC, ash content, volatile matter) from noisy documents
Automate trade certificate validation against Malaysian Rubber Board standards
Reference links and supporting material
Multilingual rubber procurement dataset containing receipts, trade permits, and technical spec sheets in Malay, Tamil, and English.
Challenge setup
Evaluates accurate multilingual field extraction and technical compliance verification.
Challenge input
Text or JSON input simulating OCR document in Malay/Tamil
Claude Agents SDK
Required for extended reasoning and structured document understanding.
Sarvam AI
Required for localized Asian language processing.
Evaluated output
JSON containing standard SMR rubber specifications and compliance boolean
- Extracted specifications must match standard schema.
- Quality score for domain-specific technical terminology translation • target: 0.88 • range: 0-1
- Benchmark: DocVQA-Multilingual
- Translation Bleu Score target: 0.88
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- E2B · Optional
- Sarvam AI · Required
- Zed · Optional
- E2B · Optional
Evaluation contract
- extraction_precision · Weight 1
- translation_bleu_score · Weight 1
Metrics
- translation_bleu_score · Target 0.88
Benchmarks
- DocVQA-Multilingual
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: Text or JSON input simulating OCR document in Malay/Tamil
Output: JSON containing standard SMR rubber specifications and compliance boolean
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Framework
- Sarvam AI · Indic AI & Translation Platform
Recipe freeze marker
No recipe freeze time is recorded.