Digital Green

Portuguese version available here.
1. Organization & Background

Digital Green is strengthening FarmerChat's AI-powered advisory systems by improving the quality, relevance, and usability of agricultural content across multiple geographies and languages. This work includes both Reinforcement Learning from Human Feedback (RLHF) activities and content development, review, and curation activities to support farmers, extension workers, and advisory systems.

The initiative seeks qualified agricultural experts who can contribute to improving AI-generated outputs, validating agricultural knowledge, and developing high-quality localized content for deployment across Digital Green platforms.

2. Objectives

The selected expert(s) will support one or more of the following areas:

A. RLHF and AI Quality Improvement

• Review and evaluate AI-generated agricultural responses.
• Improve the quality and relevance of location-specific Q&A pairs.
• Support development of Golden Q&A datasets and evaluation benchmarks.
• Provide structured feedback to improve model performance.

B. Content Development and Review

• Create, review, and validate agricultural content cards.
• Support seasonal advisory planning and content prioritization.
• Review existing content for technical accuracy, relevance, and usability.
• Contribute to localized content development across crops, livestock, climate, and regenerative agriculture topics.

3. Key Responsibilities

RLHF Activities

• Review Q&A pairs generated by AI systems, human experts, or synthetic data pipelines.
• Evaluate outputs based on agronomic correctness, relevance, clarity, trustworthiness, inclusivity, and farmer usability.
• Rank and compare alternative AI responses.
• Provide qualitative feedback to improve future model outputs.
• Participate in calibration exercises and workflow refinement sessions.
• Maintain records of completed reviews through Digital Green platforms.

Content Development & Review Activities

• Create and review content cards covering agriculture, livestock, climate resilience, regenerative agriculture, and allied sectors.
• Validate technical correctness and contextual relevance of content.
• Review content generated by AI systems and recommend edits.
• Support content adaptation for specific geographies, seasons, crops, and farming systems.
• Provide recommendations for future content priorities and improvements.

4. Areas of Expertise Sought

• Agronomy and Crop Sciences
• Livestock and Veterinary Sciences
• Agricultural Extension and Advisory Services
• Climate-Smart Agriculture
• Regenerative Agriculture
• Gender and Social Inclusion
• Soil Health and Natural Resource Management
• Horticulture and High-Value Crops

5. Deliverables

RLHF Deliverables

• Reviewed and scored Q&A pairs.
• Comparative evaluations and rankings.
• Feedback on model performance and failure cases.
• Contributions to Golden Q&A datasets.
• Monthly activity reports.

Content Deliverables

• Developed or reviewed content cards.
• Technical review comments and recommendations.
• Seasonal content plans and content prioritization inputs.
• Approved content ready for deployment.
• Monthly content review summaries.

6. Qualifications and Experience

Required

• Master's degree or higher in Agronomy, Agricultural Extension, Veterinary Sciences, Horticulture, Climate Sciences, or related disciplines.
• Minimum 5 years of experience in agricultural advisory, extension, research, or content development.
• Strong understanding of farming systems and farmer advisory services.
• Ability to review technical content critically and provide actionable feedback.

Preferred

• Experience working with AI-assisted content systems.
• Experience reviewing digital agricultural advisory content.
• Familiarity with large language models (LLMs), AI evaluation, or RLHF processes.
• Experience integrating climate and gender considerations into agricultural recommendations.

7. Engagement Model

The consultant may be engaged under one or both of the following workstreams:

Workstream A: RLHF Review and Validation

Work will be assigned through Digital Green's evaluation platform and compensated based on approved outputs.

Workstream B: Content Development and Review

Work will be assigned based on content priorities and compensated through a Level of Effort (LOE) model based on approved hours worked.

8. Compensation Structure

A. RLHF Reviewer

Expected productivity:

• Approximately 30 reviews per day.
• Minimum engagement: 20 days/month.
• Maximum engagement: 30 days/month.

B. RLHF Validator 

C. Content Development and Review

Compensation will be based on approved hours worked. Monthly hours will be agreed upon in advance based on project needs.

9. Contract Duration

Initial contract period: 6 months

The engagement may be extended based on performance, project requirements, and funding availability.

Apply by July 15, 2026.