Asks for

The ad says“Agent Orchestration: Design, deploy, and debug sophisticated agentic AI workflows, focusing on multi-step reasoning, memory integration, Model Context Protocol (MCP), and multi-agent frameworks (using LangGraph, LangChain, or equivalent setups).”
The ad says“Hands-on mastery of AWS environments and MLOps automation (Containerisation, CI/CD for ML).”
The ad says“LLM Serving & Optimisation: Deploy and operate LLM-based services directly in production, aggressively managing runtime latency, throughput, and cloud costs (using tooling like LiteLLM, vLLM, SageMaker, or Bedrock).”
The ad says“Hands-on mastery of AWS environments and MLOps automation (Containerisation, CI/CD for ML).”
The ad says“Agent Orchestration: Design, deploy, and debug sophisticated agentic AI workflows, focusing on multi-step reasoning, memory integration, Model Context Protocol (MCP), and multi-agent frameworks (using LangGraph, LangChain, or equivalent setups).”
The ad says“Agent Orchestration: Design, deploy, and debug sophisticated agentic AI workflows, focusing on multi-step reasoning, memory integration, Model Context Protocol (MCP), and multi-agent frameworks (using LangGraph, LangChain, or equivalent setups).”
The ad says“LLM Serving & Optimisation: Deploy and operate LLM-based services directly in production, aggressively managing runtime latency, throughput, and cloud costs (using tooling like LiteLLM, vLLM, SageMaker, or Bedrock).”
The ad says“LLM Serving & Optimisation: Deploy and operate LLM-based services directly in production, aggressively managing runtime latency, throughput, and cloud costs (using tooling like LiteLLM, vLLM, SageMaker, or Bedrock).”
The ad says“Traditional Grounding: Implement robust traditional machine learning models—specifically across tabular/structured data (classification, regression, ranking) or NLP—to solve complex risk and behavioral patterns.”
The ad says“Hands-on mastery of AWS environments and MLOps automation (Containerisation, CI/CD for ML).”
The ad says“Traditional Grounding: Implement robust traditional machine learning models—specifically across tabular/structured data (classification, regression, ranking) or NLP—to solve complex risk and behavioral patterns.”
The ad says“The Software Core: Outstanding proficiency in Python and deep experience working with large-scale structured data using SQL.”
The ad says“Advanced RAG Pipelines: Architect and refine retrieval-augmented generation systems, continually optimising chunking strategies, retrieval design, and end-to-end pipeline evaluation.”
The ad says“LLM Serving & Optimisation: Deploy and operate LLM-based services directly in production, aggressively managing runtime latency, throughput, and cloud costs (using tooling like LiteLLM, vLLM, SageMaker, or Bedrock).”
The ad says“The Software Core: Outstanding proficiency in Python and deep experience working with large-scale structured data using SQL.”
The ad says“LLM Serving & Optimisation: Deploy and operate LLM-based services directly in production, aggressively managing runtime latency, throughput, and cloud costs (using tooling like LiteLLM, vLLM, SageMaker, or Bedrock).”

Have you ever shipped a complex agentic AI system to production, watched it gracefully (or catastrophically) break under real-world user traffic, and spent the night refactoring its multi-step reasoning layer? If you value production-tested ML rigour over isolated Jupyter notebook prototypes, we have a system for you to build.

About TIko

Tiko is an African nonprofit committed to strengthening the potential and resilience of adolescent girls across Africa. We address the “Triple Threat” of unintended pregnancy, HIV infection, and sexual and gender-based violence by building local health ecosystems that provide stigma-free, no-cost, quality-assured services.

Our model brings together key local actors: community-based organizations (CBO) with peer mobilisers who act as health companions to girls; public and private health clinics that deliver care; and retail partners that redeem Tiko Miles -our behaviour-change incentive programme that rewards service uptake and feedback.

We invest in partners by strengthening CBO capacity, training frontline workers and providers, supporting clinic quality improvement, and compensating partners based on performance. Our technology platform connects all actors by enabling referrals, verifying service delivery, facilitating payments, and generating real-time data.

Tiko operates in six countries: Kenya, Ethiopia, Uganda, Burkina Faso, South Africa, and Nigeria, with additional offices in Portugal, the Netherlands, and the United Kingdom. For a clear overview of our work, we recommend watching this short video.

Globally, our team consists of +250 enthusiastic, international colleagues. Whether you are working from our biggest office in Nairobi, the fast-growing office in South Africa, or from home, our people are young, and our culture is global and dynamic. Our work environment is fast-paced, informal, and friendly.

For this position we will happily accept applicants from South Africa, Kenya, and Portugal.

The Job

We are looking for a T-Shaped AI / ML Engineer to join a high-ownership, agile team tackling one of our most critical scaling challenges: risk and fraud on the Tiko platform. In our engineering ecosystem, "T-shaped" isn't a buzzword- it means you bring a broad, rock-solid foundation in software engineering and data infrastructure (the horizontal bar), combined with deep, hands-on mastery of deploying modern AI and LLM-based systems in production (the vertical bar).

You will move incredibly fast to build, monitor, and optimise intelligent systems that protect our platform. This is a role where you will own entire outcomes from day one, ensuring our architectures remain scalable, secure, and performant for the communities we serve.

Key Responsibilities

Agentic Systems & GenAI Engineering (The Vertical Bar)

Agent Orchestration: Design, deploy, and debug sophisticated agentic AI workflows, focusing on multi-step reasoning, memory integration, Model Context Protocol (MCP), and multi-agent frameworks (using LangGraph, LangChain, or equivalent setups).
Advanced RAG Pipelines: Architect and refine retrieval-augmented generation systems, continually optimising chunking strategies, retrieval design, and end-to-end pipeline evaluation.
LLM Serving & Optimisation: Deploy and operate LLM-based services directly in production, aggressively managing runtime latency, throughput, and cloud costs (using tooling like LiteLLM, vLLM, SageMaker, or Bedrock).

Production Foundations & ML Rigour (The Horizontal Bar)

Traditional Grounding: Implement robust traditional machine learning models—specifically across tabular/structured data (classification, regression, ranking) or NLP—to solve complex risk and behavioral patterns.
Production Coding: Write idiomatic, clean, and production-quality Python code that seamlessly integrates into our core data and software engineering ecosystems.
Structured Data: Query, manipulate, and pipe structured data at scale using complex SQL configurations.
Iterative Systems Design: Participate in team rituals, maintain meticulous technical documentation, and champion a culture of shared learning and graceful systems degradation.

About You

You are an analytical systems thinker who understands how distributed components interact, degrade, and fail. You don't just build models; you build the evaluation and monitoring frameworks around them because you know that code is only as good as its failure modes. You are highly independent, comfortable moving fast through iterative feedback loops, and passionate about owning the actual business outcome, not just a list of technical tasks.

Requirements

Professional Track Record: 5+ years of experience operating as an ML Engineer, Data Scientist, AI Engineer, or Software Developer with a clear history of shipping code to production.
The Software Core: Outstanding proficiency in Python and deep experience working with large-scale structured data using SQL.
Production GenAI: Proven, practical experience building and serving live LLM-based or agentic systems for real users (not just internal prototypes).
ML Breadth: Solid grounding in traditional machine learning fields, with distinct depth in at least one area (tabular ML, NLP, or deep learning).
Language: Strong verbal and written English communication skills.

Bonus Points (The Differentiators):

Direct domain experience in Risk, Fraud, Credit, Insurance, Cybersecurity, or Fintech ecosystems.
Experience with GenAI observability, drift detection, and evaluation frameworks (e.g., LangSmith, Evidently, RAGAS).
Hands-on mastery of AWS environments and MLOps automation (Containerisation, CI/CD for ML).
A track record of mentoring juniors or contributing to the wider community through technical blogging or public speaking.

Recruitment Process

Introductory call with recruiter | First Interview with Tech Team | Technical Challenge | Final interview with the Tech Team

Compensation & Benefits

The gross salary range per month for this position is:

South Africa: R72,734 - R121,223 (x12)

Portugal: €3,025 - €5,041 (x14)

Kenya: Ksh354,096 - Ksh590,160 (x12)

Your final salary will be determined based on your experience and alignment with your future colleagues.

In addition to your monthly salary, we offer you:

Benefits and allowances tailored to your location.
Flexible work arrangements, including remote or hybrid options.
A personal development budget of €500 per year to invest in your professional growth
Unlimited holiday days to use as you see fit - just coordinate with your team and take the time you need to recharge.
The opportunity to shape a growing, impactful product and leave your mark on how we work
A culture built on trust - we believe you’ll do your best without the need for unnecessary rules or micromanagement

The Details

Interested? Click Apply for This Job! Want more information? Check out our website Tiko – Do more with Tiko. We only accept applications through the apply links, not by email.

Important Recruitment Fraud Alert
Please be aware that Tiko maintains a professional and ethical recruitment process.

Zero Fees: We do not charge candidates any fees for applications, interviews, or processing at any stage of the hiring journey.
Fraud Prevention: If you receive a request for money, bank details, or "onboarding equipment" payments from someone claiming to represent us, this is not a legitimate request.
Official Channels: We only communicate through our verified company email domain (@tiko.org).

If you are approached by anyone asking for payment in our name, please ignore the request.

—

Tiko prioritizes integrity in our workplace and respects your privacy.

Tiko is committed to preventing any type of unwanted behaviour by its employees at work, including sexual harassment, exploitation and abuse, lack of integrity and financial misconduct. This is why we will do reference and background screening checks on successful candidates before hiring. Tiko also participates in the Inter Agency Misconduct Disclosure Scheme. As part of this scheme, we will request information from your previous employers about any findings of sexual exploitation, sexual abuse and/or sexual harassment during your employment, or incidents under investigation when you left employment. By applying for this position, you confirm you have read and understood these recruitment procedures.

We value your privacy and understand the importance of safeguarding your personal data. We invite you to review our privacy notice for the recruitment process to understand how we collect, use, and protect your personal data during the recruitment process. Click here to view the document. By applying for this position, you acknowledge that you have read and understood our privacy notice.

Asks for

Values