Research and Development Project Assistant in Artificial Intelligence

Key information

• Position title: Research and Development Project Assistant in Artificial Intelligence
• Project: Bridging Agroecology and AI
• Organization: CREATES
• Duty station: Ngaparou (Mbour department), Senegal
• Reports to: CREATES Management
• Contract type: Fixed-term contract, 9 months, renewable subject to funding
• Level: Junior, with progressive autonomy
• Start date: As soon as possible

1. Context

CREATES is developing an exploratory project at the intersection of artificial intelligence, agroecology, participatory research and the valorization of local knowledge. The project aims to contribute to the development of a prototype AI assistant capable of supporting the dissemination of contextualized agroecological knowledge, in French and in local languages, based on a structured and verifiable documentary base.

The position is part of an action-research approach. The objective is not only to produce a technical demonstration, but to contribute to a tool that is useful, reliable, understandable and appropriable by field actors. Particular attention will be paid to the quality of sources, the traceability of answers, linguistic issues, possible biases and the sovereignty of knowledge.

2. Main mission

The recruited person will support the team in the design, development, testing and documentation of a prototype AI assistant applied to agroecology. They will contribute in particular to setting up a Retrieval-Augmented Generation (RAG) architecture, structuring document corpora, using AI-assisted development tools, and progressively adapting the system to Senegalese usage contexts and local languages.

3. Main responsibilities

AI development and prototyping

• Contribute to the development of a conversational assistant prototype based on language models.
• Help set up a simple, robust and documented RAG architecture.
• Test different methods of indexing, document retrieval and sourced answer generation.
• Identify technical problems, recurring errors and avenues for improvement.

Structuring the knowledge base

• Support the collection, cleaning and organization of scientific, technical, institutional and field documents.
• Contribute to the preparation of corpora usable by the AI system.
• Document sources, versions, limitations and structuring choices.
• Ensure that generated answers can be linked to the documents used.

Localization and local languages

• Help structure glossaries and agroecological terms in French and, as needed, in Wolof, Pulaar or other relevant languages.
• Test the effects of translation, linguistic variation and local phrasing on answer quality.
• Work with resource people to improve the linguistic and cultural relevance of the system.

Evaluation and reliability

• Contribute to the creation of test question sets and usage scenarios.
• Compare generated answers with available sources.
• Document hallucinations, errors, ambiguities and limitations of the prototype.
• Propose simple mechanisms to improve the reliability, caution and clarity of answers.

Action-research support

• Contribute to the preparation of materials for co-design and user-testing workshops.
• Translate field feedback into understandable technical needs.
• Contribute to writing technical, methodological or capitalization notes.
• Work in dialogue with researchers, developers, field actors and local partners.

4. Key technical skills

The person sought does not necessarily need to master all advanced technologies. The central criterion is the ability to learn fast, to understand what they are doing, to test rigorously and to document their work properly.

Required skills

• Good command of Python and the ability to cleanly structure a software development project.
• Practical experience with AI-assisted development tools, notably Codex, Claude Code, GitHub Copilot or equivalent tools.
• Practical understanding of the principles of Retrieval-Augmented Generation: corpora, embeddings, document retrieval, generation, source citation.
• Ability to work with varied documents: PDF, Word, texts, simple tables, structured or semi-structured data.
• Routine use of Git, GitHub, the terminal and Python environments.
• Ability to interact with language model APIs and to manage keys, errors, costs and usage limits responsibly.
• Ability to develop a simple interface or API to test a prototype.
• Ability to evaluate the quality of an AI answer: accuracy, clarity, traceability, caution and usefulness for the end user.

Appreciated skills

• Experience with vector databases or similar tools: ChromaDB, FAISS, Qdrant, pgvector or equivalent.
• Knowledge of frameworks such as LangChain, LlamaIndex, Haystack or equivalent.
• Experience with FastAPI, Streamlit, Gradio, Flask or lightweight prototyping interfaces.
• Notions of databases: SQLite, PostgreSQL, JSON, CSV, dataset organization.
• Knowledge of Hugging Face, open-source models or local deployment tools such as Ollama.
• Notions of speech-to-text, audio transcription, text-to-speech or processing of low-resource languages.
• Experience with Docker or lightweight deployment on a server, cloud platform or demonstration space.

5. Research and analysis skills

• Ability to read, synthesize and transform technical or scientific documents into usable materials.
• Interest in agroecology, food systems, local knowledge and territorial transitions.
• Sensitivity to power issues related to data, knowledge, languages and technologies.
• Ability to distinguish what is technically possible, reliable, risky or premature.
• Rigor in documentation, verification and the formulation of limitations.

6. Profile sought

The position is aimed at a person at the start of their career, finishing their studies or recently graduated, located at the intersection of software development, applied artificial intelligence and research. The ideal profile is not necessarily an expert, but must be technically solid, curious, methodical and able to work on a still-exploratory project.

Desired education

• Computer science, data science, artificial intelligence or a related field.
• An advanced degree is not essential if the person has a portfolio, GitHub projects, prototypes or concrete experiences demonstrating their skills.

Desired experience

• 1 to 3 years of professional experience.
• Concrete Python programming experience.
• Participation in at least one project using language models, a chatbot, a RAG system or an AI application.
• Experience with or strong interest in African, agricultural, multilingual or participatory contexts.

Languages

• Command of French and Wolof essential (oral and written for French, oral for Wolof), for work with field actors and the structuring of glossaries in local languages.
• Knowledge of Diola would be a significant asset.
• A good level of English is desired (AI technical documentation, tools, publications and international exchanges).

7. Expected qualities

• Progressive autonomy and rapid learning ability.
• Critical thinking toward AI tools and automatically generated answers.
• Rigor, patience and a taste for the technical details that determine the reliability of a prototype.
• Ability to work with very different profiles: researchers, developers, farmers, local organizations and institutional partners.
• Technical humility: knowing how to test, verify, ask, document and correct.
• Intercultural sensitivity and attention to real uses, beyond the technological demonstration.

8. Deliverables

• Functional prototype of an AI assistant or chatbot.
• Structured and documented documentary base.
• Simple ingestion and document-retrieval pipeline.
• A series of tests on answer quality.
• Technical documentation accessible to the team.
• Notes on the limitations, risks and avenues for improvement of the system.

9. How to apply

The application file must include:

• A cover letter (2 pages maximum) specifying the interest in the position and availability;
• A detailed curriculum vitae;
• A portfolio, links to GitHub projects or examples of work (prototypes, AI applications);
• The contact details of two to three referees.

To apply, use the Apply button below. You will receive a confirmation email after submitting your application.

Only shortlisted candidates will be contacted for an interview. CREATES encourages applications from women and from candidates reflecting the social and geographic diversity of Senegal.