AI Engineer

Company Description Gunpowder Innovations builds custom applications, web platforms, and end-to-end digital solutions that help organizations grow and compete online. The team combines smart product design with advanced technologies to turn client ideas into practical, scalable tools. By focusing on usability, performance, and business impact, Gunpowder Innovations supports digital transformation across diverse industries. Team members collaborate closely with clients, emphasizing innovation, reliability, and measurable results.About the roleWe’re building AI systems for clients who need more than a chatbot bolted onto an API. This role is for someone who can go deep: fine-tuning models, deploying open-source LLMs, building agentic systems, and designing the infrastructure that makes AI products actually work in production.Key ResponsibilitiesDesign and own the architecture for AI-powered features across client projects, from model selection through to production deploymentBuild agentic systems: tool use, multi-step reasoning, orchestration frameworks, and the harnesses that let models act reliably in real environmentsWork hands-on with open-source models, including fine-tuning, quantisation, evaluation, and self-hosting where it makes senseBuild the infrastructure around AI systems: retrieval pipelines, vector stores, inference optimisation, monitoring and evalsMake the call on when to use a foundation model API versus when to fine-tune or self-host, and be able to justify itCollaborate directly with clients to translate ambiguous problems into working AI systemsStay ahead of the open-source and agentic tooling ecosystem so Gunpowder’s clients get informed, current recommendations, not last year’s stackRequirementsReal experience building agentic systems, fine tuning models and opensource work, not just prompting a hosted APIHands-on experience building agentic workflows: designing harnesses, tool-calling logic, and evaluation loops for agents that operate with real autonomyStrong grasp of the open-source LLM landscape: tools, frameworks, trade-offs, licensingComfortable owning technical decisions end-to-end, from architecture to deploymentPractical engineering skills: Python, ML infra (Docker, cloud deployment, vector DBs), and the judgement to know what’s overkill and what isn’tA builder’s instinct: you’d rather ship something real than write another slide about AI strategy