Senior Software Engineer, AI Systems
Accra · Greater Accra
Apply directly with the employer. Mzlynn does not collect applications.
About the job
What You Will Do Build LLM-powered features into production systems — enrichment pipelines, intelligent triage, document processing, and workflow automation — using Anthropic Claude, OpenAI, and Google Gemini models from Python and TypeScript services. Design and maintain evaluation harnesses: golden datasets, scoring rubrics, and regression suites that catch quality drift before users do. AI systems are non-deterministic; you will test them like it. Engineer for cost and latency as first-class constraints: model tiering, prompt and context budgeting, caching, and batching — and defend the trade-offs with measurements. Build the guardrails production AI requires: structured-output enforcement, input validation and injection defense, output verification, and human-in-the-loop checkpoints for consequential actions. Build retrieval pipelines (embeddings, chunking, hybrid search, reranking) where the problem genuinely calls for them — and push back when it doesn’t. Own the failure modes: retries with graceful degradation, hallucination containment, confidence scoring, and feedback loops that make the system measurably better over time. What We’re Looking For 4+ years of professional software engineering experience in Python or TypeScript building production systems — this is an engineering role first, an AI role second. You have shipped at least one LLM-powered feature to real users and lived with the consequences. Be ready to tell us what broke. You understand non-determinism: you evaluate AI systems with eval suites and measurements, not vibes and demos. You can reason about cost, latency, and quality trade-offs across models and explain why a cheaper model with better prompting beats a bigger one for a given job. Strong API, queue, and database fundamentals (REST, message queues, SQL) — AI features live inside real systems. Skeptical of hype, rigorous about evidence. You measure before you claim, and your English is strong enough to defend the measurement to an international engineering team.