AI Solutions Architect

Code for Africa (CfA) uses technology and #OpenData to empower citizens. We give citizens actionable information for better-informed decision making and digital tools to amplify their voices, so that they can hold the authorities (both governmental and corporate) to account.
AI Solutions Architect


Position Summary

You’ll design, ship and maintain AI-enabled tools and workflows for teams across the alliance, and set the patterns others build on. This is a hands-on role: you’ll write and review code, configure integrations, test with users and support what you ship, not just produce designs or recommendations.

Your first priority will initially be Business Development (proposal and partner-intelligence workflows). From there, the work expands to research, editorial, investigative and programme operations. You’ll build on tools teams already use (Google Workspace, Slack, Airtable, CRMs) and on in-house capacity: DataLab for data engineering and scraping, the iLAB for AI-driven forensics, CivicSignal for NLP, and the wider AI Sandbox team for shared model access and infrastructure. You won’t be building the whole stack alone.

Why Join

Be the founding hire: shape how the alliance uses AI, from the first build to the standards other teams adopt.
Real users from day one: starting with the Business Development team, then forensic researchers, journalists, and programme staff across Africa and beyond.
Real ownership, real support: you’ll own design decisions within agreed architecture and security boundaries, backed by in-house data and AI teams, so you’re not starting from zero.

What You Will Do

Ship tools, workflows and lightweight interfaces end to end: take them from discovery and design through build, evaluation, rollout, training, monitoring and handoff.
Pick the simplest thing that works: use models or agents where interpreting language or handling ambiguity adds value, and plain deterministic code or rules for anything that runs often, must give the same answer every time, or needs an audit trail. Often that means using an AI coding agent to build the automation, not to be it.
Treat AI output as unverified until checked: build evaluations, source traceability and automated checks, and put a person where the decision matters. Assume outside content may be wrong or may try to steer the model.
Build the function, not just the tools: set up an intake and prioritisation process, a reusable library of templates, connectors and review components, and clear rules for what stays central and what programmes own.
Keep it maintainable: use version control and code review, write documentation others can run from, and build on shared Plus and AI Sandbox platforms and security standards, so everything works with the rest of Plus’s systems.
Keep costs visible: estimate running costs, compare managed, open-source and in-house options on cost, security, portability and maintenance, and replace expensive SaaS where there’s a clear case.

What This Role Is Not

This is not primarily a chatbot, dashboard or research role, and it’s not an on-demand automation desk: you’ll prioritise work with the CTO and programme leads based on value, risk and potential for reuse. You won’t own data engineering, model infrastructure or organisation-wide AI policy, but you’ll work closely with the teams that do.

What You Will Build in Year One

Four to five substantive production tools or workflows in regular operational use, starting with Business Development, each with monitoring and a named owner.
An intake and prioritisation process open to all programmes, with a clear way to select, defer and decline requests.
A Plus-wide pattern and guardrail library adopted by at least two teams.
A recommendation, based on real demand, for how the hub should grow, including the shape of a team you could eventually lead.

Required

A degree in computer science, software engineering, information systems, data science or a related field, or equivalent practical experience.
3+ years building and shipping internal tools, automation or AI-enabled workflows that people rely on regularly (not just prototypes or demos), and that you maintained after launch. You’re comfortable going from ambiguity to a working system.
Hands-on experience building with hosted model APIs and open-weight models, including structured outputs, tool use, agents, and MCP or similar integration protocols. You’re also comfortable with AI coding and workflow tools such as Claude Cowork, Claude Code or Codex, and know when not to use them.
Integration experience with several business systems, such as Google Workspace, Slack, Airtable, CRMs or email, including APIs, webhooks, OAuth, scoped credentials and rate limits. You default to live connections, and know how to stop any cached or synced copy from being mistaken for current data.
Practical AI and data risk awareness: catching outputs that are confident but wrong, handling prompt injection, protecting PII and confidential data, and matching human review to the cost of an error.
The ability to map how work actually happens, design around it with non-technical colleagues, and get them to adopt what you build, including writing guides they can actually follow.
Engineering discipline: Git, code review, tests where they matter, and clear documentation.

Preferred

A postgraduate qualification in artificial intelligence, machine learning or a related field.
Experience in non-profit, media, civic tech or international development, including proposal or partner-management workflows.
Document-heavy pipelines: parsing PDFs and DOCX files, structured extraction, and retrieval over knowledge bases that change often.
Formal evaluation practice: test sets, regression evals, red-teaming and output validation.
Experience with managed model services and gateways, such as Amazon Bedrock, OpenRouter or equivalents.
Cost modelling for LLM workloads, and security basics such as secrets management and vendor security review.
Having set up shared standards or components that other teams or organisations adopted.