1. Background

Shaqodoon is a not-for-profit organization working to improve access to information and services for communities in the Horn of Africa. As part of its commitment to enhancing communication, accountability, and early warning systems, Shaqodoon seeks to develop an automated solution for transcribing and translating Somali audio content—including Maxaa and Maay dialects—into English. This system will support humanitarian programming by enabling timely analysis of beneficiary feedback, radio call-ins, and life-saving hazard reports that could derail delivery of donor-funded programs.

2. Objective

working with Shaqodoon technology unit design, develop, and deploy an end-to-end Somali Automatic Speech Recognition (ASR) and translation pipeline that accurately converts Somali audio into Somali and English text, with built-in tools for review, continuous improvement, and integration into Shaqodoon system including early warning and CFM systems.

3. Scope of Work

The consultant/firm will be responsible for:

● Designing a modular ASR and translation pipeline.

● Integrating external APIs (e.g., ElevenLabs, OpenAI, or LLaMA) for transcription and translation.

● Building data pre-processing and post-processing layers.

● Creating a feedback interface to support human-in-the-loop validation.

● Ensuring secure storage of data and integration with Shaqodoon’s early warning systems.

● Testing the system for functionality, robustness, and accuracy under varied audio conditions.

● Enabling future replacement of third-party APIs with proprietary models.

● Providing documentation, technical handover, and staff training.

4. Key Deliverables

● Technical architecture and deployment documentation.

● MVP of the ASR and translation system processing a minimum of 200 audio files per hour.

● Feedback interface allowing manual correction and annotation of flagged transcripts.

● Metrics dashboard tracking WER, translation confidence, and human intervention rate.

● Report on WER performance across Maxaa and Maay dialects (80% accracy target).

● Integration plan with Shaqodoon’s Saadaal early warning system.

● Final training and handover package for Shaqodoon technical staff.

5. Duration

The consultancy will last 30 days, with a functional prototype delivered by the end of day 25 with remaining 5 days to be allocated for testing and integrations.

6. Required Qualifications

● Demonstrated experience developing ASR or NLP systems in low-resource languages.

● Familiarity with Somali dialects and linguistic diversity is an advantage.

● Strong command of cloud-based or containerized deployments (Docker/Kubernetes).

● Track record in building human feedback loops and secure NLP pipelines.

● Knowledge of ethical AI, bias mitigation, and data privacy.

7. Ethical and Technical Standards

● Ensure PII anonymization and privacy compliance throughout the pipeline.

● Design bias-aware models that perform across gender, dialect, and speaker profiles.

● Use open-source datasets where available and ensure all training complies with ethical data use standards.

8. Stakeholder Engagement

● Collaborate with Shaqodoon program and Technology for Development Unit teams for user feedback validation.

● Engage Shaqodoon team real-world testing.

● Include mechanisms for local reviewer participation in annotation tasks

9. Reporting

The consultant/firm will report to the Technology for Development team at Shaqodoon. Weekly check-ins and milestone-based reporting are expected.

10. Budget

Interested applicants should submit a detailed narrative and financial proposal. The total budget should reflect scalability and cost-efficiency, with phased API usage estimates depending on demand.

https://drive.google.com/file/d/1Qt2VT0xu8WMQG2nZiMXy2-KKnwbAwQlI/view?usp=sharing

How to Apply

Please send your proposal by email to [email protected] with the subject line: “Call for Proposals – Somali ASR and Translation Solution”

Interested parties are to must submit the following:

1. Technical proposal outlining the proposed solution, approach, and methodology

2. Proposed budget for project implementation

3. Detailed work plan for this project

4. Company/business profile or Consultant CV

If selected and before contract is signed the consultant/firm are expected to submit the following documentation and sign Shaqodoon Adult and Child safeguarding policies.

5. Valid business license

6. Passport copy of company director/Manager