AI Engineer (Python, ML & LLM) Johannesburg
Sompisi Calibre Solutions · Northcliff, Johannesburg, Gauteng
Overview
A fixed term AI Engineer role in Johannesburg for someone with 3 to 6 years of software experience. You will build the intelligence layer of a healthcare recruitment platform using Python, LLMs, vector search and in house models. This is hands on production AI work, not notebook research, with strong exposure to AWS Bedrock and end to end model deployment.
About the role
Our client is looking for an AI Engineer to build the intelligence layer of their platform. The systems you build will power intelligent automation across recruitment practices, CVs, documents and processes in the healthcare industry. This is a fixed term, full time role based in Johannesburg.
You will join a small team and work closely with full stack and platform engineers. Expect to own AI features from design through to production monitoring, and to make real decisions about which models and providers to use.
- Build and run Python AI and ML services that score and match workers to jobs.
- Design evaluation harnesses, guardrails and monitoring for production AI.
- Improve retrieval and matching pipelines using embeddings, vector search and taxonomy driven scoring.
- Integrate large language models for extraction, summarisation and classification.
- Train and fine tune in house models where they outperform third party APIs.
- Own data quality for AI features so that scores stay meaningful.
- Work with full stack engineers to expose AI through clean APIs.
- Monitor accuracy, latency and cost in production and iterate.
Requirements
- 3 to 6 years building software with strong, production grade Python.
- Hands on experience shipping ML, NLP or LLM powered features, not only research or notebooks.
- Ability to design and implement best practices for AI systems: evaluation, guardrails, monitoring, reproducibility and cost control.
- Practical understanding of embeddings, semantic and vector search, and retrieval augmented generation (RAG).
- Experience integrating LLMs from any provider such as Anthropic, OpenAI, AWS Bedrock or open models into real applications.
- Ability to build, train and fine tune in house models: dataset preparation, training, evaluation and serving.
- Comfortable building APIs and working with relational data, using FastAPI and PostgreSQL or similar tools.
- Solid grasp of prompt design, evaluation and the failure modes of LLM and scoring systems.
- Nice to have: familiarity with AWS Bedrock, model training frameworks like PyTorch or TensorFlow, vector databases like pgvector, agentic systems, Airflow or Prefect, MLOps, Kubernetes, and domain experience in healthcare, recruitment or HR tech.
Responsibilities
- Build and operate Python AI and ML services that score and match workers to jobs and surface results in the product.
- Design and implement best practices for production AI including evaluation harnesses, guardrails, monitoring, reproducibility and cost control.
- Design and improve retrieval and matching pipelines using embeddings, vector search and taxonomy driven scoring.
- Integrate large language models for extraction, summarisation, classification and intelligent automation.
- Build and train in house models where they outperform or reduce reliance on third party APIs, covering data preparation, training, fine tuning, evaluation and deployment.
- Own data quality for AI features, ensuring embedding and taxonomy data is correct so scores are meaningful.
- Partner with full stack engineers to expose AI capabilities through clean APIs and with platform engineers on deployment and cost.
- Monitor accuracy, latency and cost in production and iterate based on real outcomes.
Job details
- Organisation
- Sompisi Calibre Solutions
- Location
- Northcliff, Johannesburg, Gauteng
- Job type
- Contract
- Experience level
- Mid (3+ years)
- Category
- IT & Software Development
- Application Method
- Apply online
Success Guide
This is a great fit if you have a few years of Python under your belt and you have actually shipped ML or LLM features to real users, not just built models in notebooks. When you apply, highlight one or two production AI projects and be ready to talk about how you evaluated them and kept costs in check.
For the interview, expect questions on embeddings, retrieval and how you handle LLM failure modes. Brush up on your FastAPI and PostgreSQL basics too, because they build APIs and work with relational data daily. Show that you can reason about why a model behaves a certain way in production.
Since the role is in healthcare recruitment, a quick read on how CV matching works will help you stand out. Keep your CV focused on measurable outcomes and the tools you used.
How we help you prepare
This employer takes applications on their own website, so you'll submit this one yourself — but SPAAN NOW gets everything ready for you first. Here's how it works:
- Our AI tailors your CV to match this job's keywords and requirements, so it stands out to recruiters and screening systems.
- A custom cover letter is written specifically for this role — no generic templates.
- We gather the documents this job needs: CV. SPAAN NOW pulls these from your profile and packages them together, ready to attach.
- Once your application is prepared, we show you exactly how and where to apply — just follow the link to the employer and submit your ready-made CV, cover letter and documents.
- Mark it as applied under Applications to keep track, and if you're shortlisted, use AI Interview Prep to practise questions commonly asked for roles like this.
We do the hard part — tailoring and gathering everything — so applying yourself only takes a minute.
Spaan Talk
2 job seekers are in this Circle