Open to opportunities

Rafael Sapere

@rafaelsapere

Message

Lead AI Engineer specialising in production LLM agent systems, orchestration, and scalable delivery.

South Africa

Message

What I'm looking for

I’m looking to build production LLM agent systems—multi-model orchestration and LangGraph workflows—with evaluation, observability, and guardrails, shipping backend APIs that scale reliably for users.

I’m a Lead AI Engineer specialising in production LLM agent systems. I was promoted from DevOps to leading a 4-engineer AI team at ProboTalent in under six months.

I design multi-model orchestration and graph-based agent workflows, using LangGraph for multi-step reasoning, branching, and stateful execution. By redesigning generation → validation → conversion → refinement pipelines, I cut total assessment generation time by ~50% (6–10 min → 3–4 min).

I build prompt frameworks for JSON-structured outputs, schema validation, and hallucination minimisation. My approach includes guardrails for prompt-injection defence and PII handling, plus tool/function calling (MCP) and agent memory patterns.

I integrate LLM systems into production APIs with logging, evaluation, observability, and token-usage tracking. I’m currently scaling delivery to 10,000 users, while mentoring teams on LLM engineering, Python, and DevOps best practices.

Experience

Work history, roles, and key accomplishments

Current

Lead AI Engineer

Current

ProboTalent (CompRSA)

Jun 2025 - Present (1 year 1 month)

Leads a team of 4 engineers delivering production LLM agent systems, focusing on multi-model orchestration, evaluation, observability, and guardrails. Reduced assessment generation time by ~50% by redesigning multi-model orchestration pipelines and deploying the systems to production APIs scaling to 10,000 users.

LangGraph Agent Memory Patterns Prompt Engineering Evaluation Observability GuardRails Token Usage Tracking

Python Engineer

ProboTalent (CompRSA)

Jan 2025 - Jun 2025 (5 months)

Designed and implemented RAG pipelines and retrieval systems using embeddings and vector databases. Built LLM-powered applications with LangChain, including assessment generation and structured-output workflows, supported by backend services.

Python RAG Embeddings Vector Databases Semantic Search Similarity Search Reranking LLMs (LangChain Structured Outputs Backend Services

DevOps Engineer

ProboTalent (CompRSA)

Dec 2024 - Jan 2025 (1 month)

Built and optimized CI/CD pipelines and automated infrastructure management. Deployed cloud-based solutions to improve scalability and reliability.

CI CD Pipelines Infrastructure Automation cloud deployment Scalability Reliability

Network & DevOps Engineer

Jendamark Automation

Dec 2023 - Dec 2024 (1 year)

Automated network setup and AWS infrastructure using Python, Docker, and Kubernetes, scaling cloud resources. Built and maintained CI/CD pipelines (C#/.NET) and implemented security automation for threat detection and response.