Software Engineer — AI

About the role

CognitiveClouds is looking for software engineers who want to build AI into real products.

You will design and ship AI-enabled systems for customers across different industries, working on problems that may involve agents, retrieval, structured and unstructured data, model orchestration, evaluation and human-in-the-loop workflows.

This is an engineering role, not a prompt-writing role. The difficult work is building the software around the models: understanding the problem, designing the architecture, connecting enterprise systems, evaluating behaviour, handling failure modes and turning promising prototypes into dependable products.

You will work closely with product managers, designers, engineers and customers from early discovery through production.

What you'll do

• Design, build and deploy AI-enabled product capabilities.
• Build agentic and tool-using workflows where they genuinely improve the product.
• Integrate foundation models from multiple providers.
• Build retrieval, context and knowledge systems.
• Design structured-output and tool-calling workflows.
• Develop evaluation frameworks for AI quality, reliability and safety.
• Build backend services and APIs around AI capabilities.
• Work with frontend engineers to create intuitive human-AI experiences.
• Prototype rapidly and test ideas against real user problems.
• Improve latency, reliability, observability and cost of production AI systems.
• Work with customer data and existing enterprise systems.
• Turn successful project patterns into reusable CognitiveClouds capabilities.
• Stay current with new models and techniques through hands-on experimentation.

You might be a fit if

• You are first and foremost a strong software engineer.
• You have built production applications involving LLMs or modern ML systems.
• You understand agents, retrieval, embeddings, tool use and model APIs.
• You are comfortable with Python and/or TypeScript.
• You understand APIs, databases, distributed systems and cloud infrastructure.
• You know that a successful AI product requires far more than calling a model.
• You can move comfortably between experimentation and disciplined engineering.
• You make sensible choices about when AI should — and should not — be used.
• You enjoy ambiguous technical problems and take ownership through deployment.