- Scrums.com is looking for a Lead AI Engineer, Voice AI and LLM Systems, to help shape their client's next-generation voice and LLM platform for contact centre and collections operations.
- This is a hands-on leadership position. You'll help lead a small team while owning live voice agents and speech pipelines that real customers rely on at high call volumes.
- You'll take systems that already work in production and scale them to new clients, languages and channels, including email and WhatsApp.
- You'll work directly with the Head of AI on architecture and priorities, with real influence over how the platform grows.
- It suits a high-agency engineer who spots what needs doing, takes ownership and ships without waiting to be told.
- The role is full time and onsite in Sandton, Johannesburg, on a 12-month contract starting December 2026.
Responsibilities
- Build and improve real-time voice agents, including prompts, tool calling, webhooks and telephony integration.
- Own the speech and LLM pipelines: transcription, diarisation, voice activity detection (VAD) and LLM-based call scoring.
- Build and maintain evaluation suites, including LLM-as-judge and regression tests running in CI.
- Improve latency, reliability and cost efficiency, including GPU and LLM spend, failure handling and observability.
- Ship production Python services on Azure (Functions, Kubernetes, queues and SQL).
- Help build new channels: email and WhatsApp agents, post-call summaries and outcome scoring.
- Work with the Head of AI on architecture, priorities and production risk.
- Mentor engineers and review their work while staying hands-on.
Experience and education
- 6+ years of software engineering experience, including at least 2 years building LLM or speech AI systems in production.
- A track record of shipping AI products that real users and businesses depend on, beyond prototypes.
- Experience leading or mentoring engineers while remaining hands-on.
- Strong Python, with hands-on experience of LLM platforms and model serving (Azure OpenAI, OpenAI, vLLM).
- Experience with speech tooling: speech-to-text, text-to-speech, Whisper, diarisation and VAD.
- Experience running services on Azure, with Docker, Kubernetes, CI/CD (Azure DevOps) and observability (Application Insights, Grafana).
- Based in, or able to commute daily to, Sandton, Johannesburg for full-time onsite work.
- Advantageous: contact centre, collections, fintech or other regulated environments.
- Preferred: a degree in Computer Science, Engineering or similar, although a proven production track record matters more.


