Senior AI Engineer - Voice AI & LLM Systems (Hybrid, Sandton)

Introduction

Ever wanted to build AI that real users depend on, not a proof of concept? At Nutun we build production AI products for contact-centre and debt-collection operations: realtime voice agents, automated call QA (speech + LLM scoring), and a client-facing platform portal that gives clients live visibility into AI performance and self-service control over their campaigns. Text-based agents for email and WhatsApp are next.

We're looking for a Senior AI Engineer to work on Zoey . You'll be the engineer behind her: LLM orchestration, speech-to-text and text-to-speech, evals, and latency and cost at scale. It's a high-autonomy role in a small, agile team led by the Head of AI. We lead by building, not by slide deck, and AI coding tools like Cursor and Claude are encouraged.

Location: Sandton, Johannesburg (hybrid flexibility) : Employment type: Permanent

Duties & Responsibilities

• Build and harden realtime voice agents: modular prompts, multi-assistant flows, tool-calling, webhooks and telephony/operational concerns

• Improve automated QA pipelines: VAD, diarization, language gating, transcription, alignment/attribution and LLM scorecard scoring

• Implement structured LLM pipelines: schemas, batching, failure handling and score recalculation

• Own features in Azure Functions / Kubernetes workers: queue consumers, artifact pipelines and SQL integrations

• Contribute to post-call intelligence features such as outcome scoring, summary generation and structured data extraction

• Expand and maintain evaluation suites: scenarios, LLM-as-judge and CI regression

• Contribute to email/WhatsApp agent work, reusing tools, outcomes and observability

• Review code, unblock Intermediate and Junior engineers, and improve runbooks and operational quality

• Use AI coding tools (e.g. Cursor, Claude) as a normal part of day-to-day delivery

Desired Experience & Qualification

• 5+ years' engineering experience, including 2+ years of applied LLM / speech AI in production

• Strong Python (APIs, typing/validation, testing, packaging)

• OpenAI / Azure OpenAI experience (chat, structured outputs, batch patterns)

• Azure messaging, storage and SQL-backed services

• Proven ownership of ambiguous, multi-system features

• Proof you've shipped something real users depend on

Advantageous (voice AI, collections or fintech experience is a bonus, not a barrier):

• Experience with Cursor, Claude or similar AI coding assistants

• Voice agent stacks: latency, barge-in, endpointing awareness

• Speech ML in production: PyTorch, Whisper, diarization, ONNX, GPU Docker

• vLLM / guided decoding and Kubernetes worker patterns

• Contact-centre QA, collections or fintech domain experience

• SQLAlchemy / Azure SQL; IaC awareness or strong infra collaboration

Tech stack: Python, Azure, Azure OpenAI, OpenAI, Kubernetes, SQL, Azure SQL, Whisper, vLLM, ONNX, PyTorch, Docker, Cursor, Claude

Interested?

If you've shipped something real users depend on, we'd like to hear from you. Apply through PNet with your CV and include a link to your GitHub profile or portfolio so we can see your work. Know someone who'd be a great fit? Pass this on.