Machine Learning Specialist

Job description

# **About the Company** Nuru Solutions is a B2B agricultural data intelligence platform operating across Kenya, Malawi, Nigeria, and Somalia. We combine satellite imagery, weather data, ground truth, and machine learning to deliver farm-level intelligence to insurers, lenders, and agribusinesses serving smallholder farmers. Our platform powers six core analytical pillars: crop health monitoring, yield prediction, risk profiling, farm boundary detection, credit risk, and market price forecasting. We achieve 80–98% accuracy through a hybrid approach that fuses multi-source satellite data, ML models, and validated ground-truth data, a combination that outperforms single-source competitors. ## **About the Role** Nuru is at an inflection point. We have validated product-market fit with, have a growing institutional pipeline, and proven model accuracy across multiple countries. As we scale from pilot delivery to commercial-grade operations, we need a senior ML leader who will own the integrity, reproducibility, and continuous improvement of every model we ship. This person will be responsible for transforming Nuru’s ML function from a talented-but-informal operation into a rigorous, scalable, and auditable system that institutional clients can rely on. ## **What You Will Do** 1. **Model Ownership & Lifecycle** - Own the complete ML lifecycle - Lead model development, training, validation, deployment, and ongoing performance monitoring for all production models. - Architect and maintain reproducible ML pipelines on AWS, ensuring all models are version-controlled, documented, and independently reproducible. - Drive multi-crop expansion (from maize to beans, sorghum, potatoes, and horticultural crops) and cross-country model generalisation across diverse agroecological zones and cropping calendars. 2. **Governance, Validation & Quality** - Own and enforce Nuru’s Model Validation Protocol, including the Test 1 / Test 2 distinction: internal holdout results (Test 1) are for internal use only; independent field validation (Test 2) is the sole metric approved for external reporting. - Execute and maintain Model Validation & Sign-Off Reports for all production models (19 models currently require individual sign-off). - Lead Quarterly Model Governance Reviews, documenting model health, drift, and accuracy trends. - Enforce the model change protocol: no model modification ships without documented justification, before/after accuracy comparisons, and sign-off. - Establish pre-delivery quality assurance for all client-facing datasets and analytics, including automated checks for data integrity issues (e.g., impossible values, distribution anomalies). 3. **Ground-Truth & Data Strategy** - Design and oversee ground-truth data collection strategies, integrating field surveys (KoboToolbox), drone imagery, crop-cut samples, and in-person validation. - Work with sparse, noisy, and incomplete ground-truth data typical of smallholder agriculture contexts, developing robust approaches to training and validation under data scarcity. - Collaborate with operations teams across Kenya, Malawi, Nigeria, and Somalia to ensure field data quality and timeliness. 4. **Team Leadership & Stakeholder Communication** - Mentor and develop junior data science team members, establishing standards for code quality, documentation, and peer review. - Collaborate with product, engineering, and client-facing teams to translate model capabilities into actionable intelligence delivered via dashboards, APIs, SMS/WhatsApp, and client reports. - Defend model methodology and accuracy claims to institutional partners, including actuaries, risk analysts, and underwriters at organisations like Swiss Re and FSD Africa. - Present technical findings clearly to non-technical stakeholders, including investors, board members, and partner executives. ## **What You Have** **Must-Haves (Required)** - 7+ years of professional experience in machine learning, with demonstrated expertise in geospatial ML, remote sensing, or agricultural applications. - Hands-on experience with satellite imagery analysis (Sentinel, Planet Labs, or similar), vegetation indices, and time-series modelling for crop or environmental applications. - Proven track record building ML governance and quality systems — ideally in environments where formal processes did not previously exist. - Strong MLOps foundation: version control (Git), model registry, experiment tracking, reproducible training pipelines, and deployment automation. - Experience managing or mentoring small technical teams (2–5 people) in fast-moving, resource-constrained environments. - Comfort working with sparse, noisy, or incomplete datasets and designing robust validation approaches under data scarcity. - Ability to communicate technical complexity clearly and credibly to institutional clients, investors, and non-technical leadership. - Self-directed problem-solver who thrives in early-stage environments where you build the systems, not just use them. **Strongly Preferred** - Understanding of agricultural systems and smallholder farming contexts in East or Southern Africa. - Experience with AWS cloud infrastructure (S3, EC2/ECS, IAM) for ML workloads. - Familiarity with insurance, credit risk, or financial product design in agricultural or development contexts. - Experience with ensemble methods (Prophet, LSTM, XGBoost), CNNs, and foundation models (SAM or similar) in production settings. - Prior work with ground-truth data collection programmes (crop cuts, field surveys, drone validation). ## **What We Offer** - A company recognised as one of the 30 most promising African startups - A validated impact: 25,000 farmers served - Direct collaboration with the CEO and a lean, mission-driven team across four countries. - The opportunity to build the ML governance and infrastructure layer for a platform that is becoming critical data infrastructure for African agrifinance.