AI Engineer (Junior/Middle)

Job Description ABOUT THE JOB

• Architect and build advanced AI agentic systems end-to-end — including planning, memory, tool use, multi-agent delegation, evaluation loops, and guardrails — always choosing the right abstraction for the real problem, not just what’s trending
• Design and implement LLM-powered applications in production, defining prompt and context strategies, tool interfaces, retrieval and reranking logic, structured outputs, streaming, and evaluation — across text and multimodal inputs (vision, documents, audio)
• Own the evaluation strategy for AI/agent systems: build offline evaluation datasets, design online LLM-as-judge loops, and set up regression harnesses that focus on metrics that truly impact users, not just dashboards
• Optimize and “squeeze” AI systems for performance and cost: implement prompt caching, batching, speculative decoding, model routing, token budget management, and latency targets; monitor and understand P50/P99 behaviour and continuously improve it
• Lead productionization of AI workloads: design APIs and inference services, build RAG/embedding pipelines, containerize workloads, and handle CPU/GPU deployment while optimizing latency, throughput, reliability, and cost
• Contribute upstream to the AI ecosystem: read SDK source code when documentation is limited, open PRs to open-source agent frameworks, and write clear bug reports for vendors when orchestration services misbehave
• Operate AI systems in production with strong AI platform & MLOps practices — model/version management, CI/CD, evaluation gates, observability, autoscaling, rollback, and failure handling, using tools such as Docker, Kubernetes/AKS, Azure AI services, vLLM or NVIDIA Triton
• Work hands-on with multimodal models (vision-language, document AI, audio) from ingest to grounded output, including OCR, layout understanding, tables, and speech processing
• Collaborate closely with Product Owners, engineers, and stakeholders to understand business needs and translate them into robust agent architectures, LLM workflows, and technical solutions
• Mentor other engineers and set the technical bar through design reviews, code reviews, and technical writing that shape how the team thinks about agents and AI systems

About You

• Bachelor’s degree in Artificial Intelligence, Computer Science, or a related field — or equivalent practical experience
• 1+ years of experience as an AI / AI Agent Engineer, working on real-world AI or LLM-based applications
• Solid machine learning fundamentals: able to clearly explain transformers (attention, positional encoding, KV cache, tokenisation, sampling) and familiar with core research papers beyond just the abstracts
• Deep LLM application experience: you have built multiple production systems on top of frontier models (e.g., Anthropic, OpenAI, Gemini, open-weight models) and understand practical edge cases such as tool-use stability, structured-output failure modes, long-context degradation, prompt-injection defence, and cost control
• Proven agent systems depth: experience building systems with real agent behaviour — planning, memory, tool orchestration, multi-step execution, and error recovery — beyond simple single-prompt loops; multi-agent coordination (delegation, sub-agent protocols, MCP-style tool servers) is a strong plus
• Hands-on multimodal experience with vision-language models, document AI (OCR, layout, tables), or audio, including end-to-end pipelines from data ingest to grounded outputs
• Strong engineering craft with Python at a senior level — async programming, typing, testing, packaging, observability — and the ability to navigate and contribute to large codebases with clean Git practices
• Solid production AI engineering background: taking ML, LLM, embedding, vision, or multimodal models from prototype to production, designing APIs and inference services, building RAG/embedding pipelines, and deploying/optimizing workloads on CPU/GPU
• Practical AI platform & MLOps experience: operating AI workloads in production with model/version management, CI/CD, evaluation gates, observability, autoscaling, rollback, and failure handling; experience with Docker, Kubernetes/AKS, Azure AI services, GPU inference, vLLM, or NVIDIA Triton is a strong plus
• Fluent with modern coding agents (Claude Code, Cursor, Copilot, or equivalents) and able to design prompts, context windows, and tool boundaries to use them effectively while understanding their limitations …