AI/ML Engineer, Pretoria

AI/ML Engineer Purpose Serve as the technical backbone of the AI Hub. Design, build, productionise and maintain core AI/ML infrastructure and reusable components that will power the DT2.0 platform build, priority use cases, and the Group's intelligent transformation.

Key Responsibilities

Design and implement core AI infrastructure— model serving, MLOps pipelines, RAG architectures, vector databases, DT Data Lake integration, rules engine and microservices.

Lead technical delivery of high-impact use cases— AI-Assisted Legacy Modernisation, Intelligent Operations, and AI-native components of DT2.0.

Build and maintain the foundational AI Agent Layer and reusable AI services across onboarding, operations and sales enablement, including multi-agent orchestration.

Apply AI/ML techniques to accelerate the DT2.0 platform build (automated code analysis, intelligent configuration generation, predictive monitoring).

Implement production-grade evaluation, observability and monitoring— eval harnesses, regression testing, A/B testing of models, continuous quality measurement.

Own LLM cost and latency optimisation— model selection, caching, prompt engineering, token economics, and routing logic.

Establish AI security standards— prompt injection defence, output validation, secrets management, model access controls — to fintech-grade requirements.

Ensure all solutions are modular, scalable, cloud‑native, secure, and fully compliant with the Responsible AI Policy, PCI‑DSS and POPIA.

Mentor squad members and contribute to AI standards and best practices across the organisation.

Track and report technical KPIs and ROI of AI initiatives.

Requirements&Qualifications Essential

BSc Computer Science or equivalent technical degree— non‑negotiable.

3–6 years software engineering experience with AI/ML in production.

Strong C#/.NET proficiency (primary API stack), shipping production APIs.

Python proficiency for AI/ML workloads.

Modern AI/ML frameworks: LangChain, LlamaIndex, Hugging Face, OpenAI/Anthropic APIs, scikit‑learn, PyTorch/TensorFlow.

RAG systems with vector DBs (Pinecone, Weaviate, pgvector, Qdrant).

Agentic / multi‑agent systems (LangGraph, CrewAI, AutoGen, custom orchestration).

MLOps, Azure (preferred), Docker/Kubernetes, data pipelines.

Azure DevOps (primary) for CI/CD.

Passion for clean, production‑grade code.

Bonus / Advantageous

FastAPI for AI service‑specific work.

Eval / observability tooling— LangSmith, Langfuse, Arize, W&B.

Fine‑tuning approaches — PEFT, LoRA, QLoRA, and trade‑off judgement.

GitHub Actions or other modern CI/CD tooling.

Modern data engineering— dbt, Spark/Databricks, Kafka.

Fintech/payments or large‑scale system modernisation background.

What Success Looks Like in 12 Months

Core AI infrastructure is live and reused across multiple DT2.0 squads.

Two priority use cases (Legacy Modernisation Toolkit and Operations layer) are in production.

Measurable acceleration of DT2.0 delivery velocity, visible to EXCO.

AI services in production with defined SLAs, eval coverage, and cost/latency baselines.

#J-18808-Ljbffr