AI Software Engineer

Were Avvoka We build drafting technology that's transforming the way the world contracts - our platform helps legal teams move faster through contracts using automation and AI, without taking judgement away from the lawyers in control. We believe technology should amplify expertise, not replace it. Avvoka is trusted by over 20% of the AmLaw 100 law firms, global banks and enterprises, and we've grown largely through product strength. With headcount and revenue contuining to scale rapidly year on year, we're now moving from a product-led path into a globally recognised legal-tech brand. We're at an inflection point: evolving how the world's most sophisticated legal teams work and building a company where thoughtful people can do the best work of their careers. Why join us This isnt add an AI chatbot to the product. Youll be building the system that builds the software. Were building The Factory: an agentic system that turns GitLab issues into production-ready merge requests automatically. Think multi-agent pipelines, LLM orchestration, and developer tooling that actually ships code. The work is hands-on and high-leverage: if The Factory gets better, Avvoka ships better and faster across the board. Youll work directly with the team lead on architecture and approach, with real autonomy to shape how AI-powered development works inside a legal tech startup. Youll also be close to a team that cares about reliability and trust: in legal workflows, almost right isnt good enough so the engineering craft (evaluation, guardrails, observability) matters as much as the model capability. Role details Department: Engineering (AI platform / developer productivity) Engagement focus: Individual contributor delivery (contract) Primary point of contact: Engineering Lead (The Factory) Location: Prague Hybrid (3 days a week in office) Billable hours: up to 160 a month Compensation: Competitive, based on experience Start date: As soon as possible What youll do Build and extend The Factory Build and extend The Factory, our multi-agent system that processes GitLab issues end-to-end through specialised agents. Ship production-grade workflows that move from issue plan code review merge request. Iterate quickly while keeping quality high through strong interfaces, tests, and system design. Orchestrate agentic workflows that are reliable Design robust agentic workflows using tools like BAML, MCP, and DSPy (or equivalents). Build guardrails that keep outputs predictable: structured outputs, tool/function calling patterns, retries, and fallbacks. Ensure workflows degrade gracefully when context is missing, requirements are ambiguous, or models behave unexpectedly. Write the glue between LLMs and real codebases Implement context retrieval across repos: ownership boundaries, relevant files, conventions, and dependencies. Build code generation and automated review loops that respect architecture and patterns in the codebase. Create merge request creation flows (branching, commit hygiene, CI awareness, and reviewer-friendly diffs). Stay AI-native and keep us ahead Work daily with AI-native dev tools: Claude Code, Codex, Gemini CLI, and whatever drops next week. Continuously evaluate new AI development tools and decide whats worth integrating (and what isnt). Improve developer experience: faster cycles, fewer regressions, better signals for humans reviewing AI-generated changes. What success looks like To ensure your application has the best opportunity of success, your CV could cover the below measures of success with quantifiable results (e.g. percentages, growth, reductions, impact) Increased issue merge request throughput (e.g. reduced cycle time, increased weekly shipped PRs/MRs, improved lead time). Improved quality and reliability of agent output (e.g. higher pass rate on eval suites, fewer CI failures, fewer reviewer-requested rewrites). Reduced engineering overhead (e.g. fewer manual steps, fewer repeated fixes, lower rework rate, improved developer satisfaction signals). What youll bring Youve actually built with AI coding/agent tools in real workflows (not just demoed them). Strong TypeScript and/or Python (bonus if youve worked with Ruby on Rails). Comfort with prompt design, agent orchestration patterns, and basic LLM evaluation (offline and/or in-product signals). You understand software architecture well enough to teach an agent about it: boundaries, trade-offs, conventions, and what good looks like in a real codebase. Hard requirement (the only one): hands-on experience with at least one AI-native dev/agent tool (e.g. Claude Code, Codex, Gemini CLI, or similar). If youve used one deeply, we can help you ramp on the rest. Bonus points if Youve built multi-agent pipelines that coordinate planning, coding, review, and integration. Youve implemented retrieval and context-building for large repos (ownership, dependencies, patterns). Youve built eval harnesses (golden sets, regression checks, rubric scoring, or CI-integrated gates). If youre excited about this role but your experience doesnt align perfectly with every item above or you havent used all the technologies mentioned we encourage you to apply anyway. We care most about strong fundamentals, curiosity, and evidence you can ship. Core attributes we value across all roles Adaptability in dynamically evolving settings A proactive, solution-focused mindset with ownership A collaborative spirit, supporting and mentoring others If youre excited about this role but your experience doesnt align perfectly with every qualification, we encourage you to apply anyway you might be just the candidate were looking for. Our hiring process CV review We review your CV for evidence of role alignment, impact, and ownership. Screening call A short call to understand your background, motivations, and what youre looking for next. Assessment interview A practical session focused on how you approach problems relevant to the role. Senior interview A deeper conversation on technical judgement, collaboration, and role fit. Meet the team Time with future teammates to ensure mutual fit and answer your questions. What we offer Clear scope of work, with clear success criteria and meaningful deliverables Ability to invoice via own company / umbrella / sole trader Autonomy over how and when work is delivered Access to necessary systems, tools, and documentation Clear success criteria and delivery milestones Opportunity to work on complex, high-impact problems Exposure to enterprise / scale-up environments Ability to shape systems, processes, or architecture Strong portfolio / reference value Equal opportunities Were committed to building an inclusive workplace where everyone feels respected, valued, and able to do their best work. We welcome applications from all backgrounds and do not discriminate on the basis of race, colour, religion, gender, gender identity or expression, sexual orientation, national origin, disability, age, or any other protected characteristic.