ABOUT YOU
You are an AI practitioner who builds things, not slides — and shapes strategy, not just tickets. You live at the intersection of product thinking and technical architecture, and you thrive there, because that is the only place where AI actually changes how a product works.
You have shipped LLM-powered systems into production and have the scars to prove it. You think in agents, not features. You stay at the frontier: when a new model or paradigm lands, you have already tried it and formed an opinion. And when you propose something, you can build a working proof of it.
THE ROLE
You will join Xsolla's Monetization Products team as the person who sits between product strategy and engineering architecture — with strong influence on both. This is not an ML research role. It is not a product management role. It is the bridge: someone who prototypes fast, gets ideas accepted, and designs systems that engineering teams can own and operate at scale.
You will architect our Agents-first approach across the product portfolio — defining how natural language, agentic control, and multi-modal interfaces layer on top of Xsolla's existing product capabilities. Critically, your designs will be built for engineering team ownership from day one: with clear handoff documentation, observable instrumentation, and runbooks that allow product engineering squads to maintain and evolve what you ship — without requiring your continued involvement in day-to-day operations.
WHAT YOU'LL DO
Design end-to-end agentic architectures — tool-use schemas, intent parsing, multi-step orchestration, and safety guardrails — engineered for long-term ownership by product engineering teams, not solo maintenance.
Define the multi-modal interface strategy across our product portfolio: how the same capability is exposed via UI, API, SDK, and agentic natural language — consistently and without duplication.
Design the horizontal LLM platform layer — shared RAG pipelines, prompt libraries, vector search infrastructure, and evaluation frameworks — that product engineering teams can build on and operate independently.
Prototype rapidly to validate AI product hypotheses before full engineering investment. Prototype acceptance by product teams is a primary success signal.
Ensure every system you architect comes with the observability, documentation, and engineering runbooks needed for a product squad to take ownership confidently.
Shape product strategy alongside Product leadership: actively influence what AI capabilities get prioritized, in what order, and with what trade-offs.
Select and govern LLM providers and deployment strategies per use case — balancing cost, latency, accuracy, and privacy requirements.
Drive alignment across Engineering, Product, and Design on what 'agent-ready' means for each product surface.
Mentor engineers on LLM integration patterns, agent evaluation, and production deployment practices — building the team's capability to own what you design.
WHAT YOU'LL NEED
Required
5+ years of engineering experience, with at least 2 years designing and deploying LLM-powered systems in production.
Proven track record designing agentic systems: tool-use, function calling, multi-step reasoning, orchestration, and error recovery at production scale.
Experience designing AI systems for engineering team ownership — including observability standards, handoff documentation, and runbooks that let other teams maintain what you build.
Hands-on experience with major LLM APIs (OpenAI, Anthropic, Google Gemini) and at least one open-source model stack.
Experience building RAG pipelines with vector databases and orchestration frameworks (LangChain, LlamaIndex, or custom).
Strong Python engineering skills — production-grade LLM services, not just notebooks.
Demonstrated ability to influence product direction: you have shaped what gets built, not just how.
Clear communication in both directions: architectural trade-offs to engineers, business outcomes to executives.
Nice to Have
Background in gaming, payments, or e-commerce — understanding of developer workflows, monetization models, or merchant operations.
Fine-tuning experience (PEFT/LoRA) for domain-specific model adaptation.
Experience with multi-agent orchestration frameworks (AutoGen, CrewAI, or custom).
Familiarity with LLM evaluation frameworks (RAGAS, DeepEval, or custom harnesses).
Exposure to EU AI Act, GDPR, or other AI compliance frameworks.
