Required: minimum requirements include:
4+ years building and shipping software, with meaningful hands-on experience building AI-powered products or systems
Fluency in Python and TypeScript
Demonstrated experience designing and building agentic AI systems: multi-step task execution, tool use, memory, planning, and error recovery
Strong context engineering instincts: you think about the full information architecture a model needs to be useful, not just how to phrase a prompt
A systematic approach to evals: you design for measurability, not just intuition, and you know how to tell whether an AI feature is actually working
Familiarity with the broader AI ecosystem: open-source tooling alongside commercial APIs and nonprofit access programmes from leading labs
Strong system design instincts around AI: you think about latency, fallbacks, cost, and reliability, not just model quality
Sound judgement on responsible AI: bias, fairness, transparency, and the limits of what a model should be asked to do
The ability to communicate clearly across the room: to an engineer debugging a pipeline and to a journalist or funder asking what it all means
Fluency in English
A degree in Computer Science, Engineering, or a related field — or equivalent experience you can point to through your work and portfolio
Preferred: candidates who are able to demonstrate the following will have an advantage:
Experience deploying open-source LLMs in production environments
Existing relationships or experience working with AI lab programmes: Anthropic for Startups/Nonprofits, OpenAI for Nonprofits, Google.org AI access, or similar
Familiarity with vector databases, embedding models, and knowledge graph approaches
Experience with multimodal AI systems
Background in containerisation and cloud infrastructure (Docker, Kubernetes, cloud-hosted model deployment)
Experience in civic technology, investigative journalism, international development, or human rights contexts
Experience with multilingual NLP, particularly for low-resource or African languages
Fluency in French, Arabic, KiSwahili, or another major African language
Experience working across international, cross-cultural technical teams
Language and Location Requirements:
Location: Fully remote — open to candidates anywhere in the world, with a preference for those based in Africa
Languages: English required; French, Arabic, KiSwahili, or any other major African language is a significant advantage
About the Role:
The role sits within the TechLab. You will collaborate with a distributed, multidisciplinary team of engineers, designers, data journalists, and product managers. You will also engage directly with external partners ranging from investigative newsrooms to human rights defenders.
CfA takes a pragmatic, portfolio approach to the AI landscape. Open-source models such as Mistral, Qwen, Gemma and others sit at the core of our infrastructure where data sovereignty and auditability matter most. But we also engage with leading AI labs through their nonprofit programmes, when frontier capability serves a specific need. Part of this role is maintaining the relationships and technical fluency to move across that landscape intelligently.
One of your early priorities will be leading the technical architecture of our AI Innovation Sandbox, a self-contained ecosystem that gives civil society organisations access to this full range of AI infrastructure and tooling. This is a flagship initiative, but it is one of many: you will be expected to identify, shape, and lead AI work across the full breadth of CfA's portfolio as the field evolves.
Responsibilities: Your work schedule will include:
Navigate the AI model landscape
Make and maintain principled decisions about when to use open-source models, when to leverage frontier models through nonprofit partnerships, and how to architect systems that avoid lock-in either way
Cultivate relationships with leading AI labs such as Anthropic, OpenAI,and others, staying close to how their technology, access programmes, and priorities are evolving
Monitor the broader ecosystem continuously, and bring the right capabilities to CfA's work before partners and peers fall behind
Engineer context, not just prompts
Design the full context that makes models useful: system instructions, retrieval strategies, memory architecture, tool outputs, conversation state, and structured reasoning chains; not just individual prompts
Build and maintain context engineering frameworks, agent templates, and workflow orchestration tools that the wider team and partner organisations can use without deep AI expertise
Create domain-specific AI assistants grounded in curated, high-quality knowledge bases, making specialist knowledge accessible and actionable at scale
Design and run evals
Build evaluation frameworks that give the team genuine confidence that AI systems are working as intended, not just anecdotally, but measurably
Treat evals as a first-class engineering discipline: defining what good looks like before building, not after
Identify failure modes proactively, particularly in African linguistic and cultural contexts where standard benchmarks often fall short
Build agent systems that do real work
Design and develop AI agents capable of planning, executing multi-step tasks, using external tools and APIs, handling errors gracefully, and operating with appropriate degrees of autonomy
Move the team beyond single-turn interactions toward systems that can reason, retrieve, act, and self-correct across longer workflows
Apply agentic thinking to how the team itself works; using AI-assisted development, automated pipelines, and agent tooling to move faster and build better across the portfolio
Build and ship AI-powered products
Design and develop AI features across CfA's platforms, from RAG systems and agentic pipelines to tool integrations and multimodal applications
Collaborate with product managers and designers from the start of a feature, not the end to turn user needs into sound technical decisions and technical possibilities into experiences people can actually use
Own the full cycle from prototype to production, including the unglamorous parts: versioning, output testing, edge case handling, and knowing when to ship and when to go back
Drive responsible AI practice
Embed bias detection, ethical review, and human rights considerations into how CfA builds and deploys AI; particularly in African linguistic, political, and social contexts
Develop clear documentation and governance protocols that ensure accountability and auditability across the portfolio
Represent CfA's AI thinking externally: in publications, partnerships, conferences, and peer networks
Build capability across the organisation and beyond
Grow CfA's internal AI literacy across technical and non-technical colleagues
Support partner organisations such newsrooms, civil society groups, and researchers, through direct technical guidance and capacity building
Stay closely connected to the global applied AI community, bringing relevant advances back into CfA's work