Applied LLM Engineer

• Own YOlanda’s system prompt, including persona, tone, safety rules, refusals, and handoff behavior
• Write and maintain prompts for V3 user actions, including plan changes, top-ups, balance queries, offers, troubleshooting, and account support
• Design conversations to reach a product or service within five responses
• Define YOlanda’s boundary of knowledge and create seamless handoffs to YOnC
• Manage context about users, accounts, and plan catalogues; break tasks into chains and tool calls
• Build and maintain bilingual evaluation sets with test cases, pass conditions, and trackable scores
• Measure the YOlanda/YOnC boundary, fact-check plan and account information, and red-team top-up and rewards flows
• Trace calls and monitor prompt versions, models, tokens, cache hits, latency, and cost
• Set latency and availability SLOs and build alerts
• Own gateway timeouts, retries, and provider/model fallbacks
• Design cacheable prompts and build semantic, retrieval, and tool-result caching with correct scoping and invalidation
• Optimize token budgets, context trimming, output length, and batch APIs
• Benchmark and route frontier and open models by task based on accuracy, latency, and cost
• Collaborate with engineers running open models on YOnC compute
• Keep prompts, models, and configuration versioned, reviewed, tested, owned, and rollback-ready in git
• Build CI release gates, canary deployments, staged rollouts, and kill switches
• Audit account- and money-related actions and redact personal data before logging or sending to third-party providers
• Implement runtime guardrails including prompt-injection defenses, tool permissions, spend limits, and human-approval points
• Set cost budgets per feature and model with alerts and translate AI/money/operator-account rules into code controls
• Open pull requests against the LangChain and LangGraph build alongside the two owning engineers
• Work daily in models and production without managing people

• Two or more years building with large language models in production, at real traffic
• Python proficiency, including scripts, API clients, data handling, and tests
• Git, JSON, regex, and Markdown
• Shipped production systems with LangChain and LangGraph; able to read and modify an unfamiliar graph safely
• Experience writing and shipping system prompts and building evaluation harnesses
• Ability to change one variable at a time and record results
• Production experience with tracing, SLOs, fallbacks, and correctly invalidated caching
• Demonstrated before-and-after improvements in cost and latency with quantified results
• Practical governance experience with versioning, release gates, audit logging, and personal data handling
• Production retrieval experience with chunking, embeddings, reranking, and grounding answers in sources; knowing when retrieval is inappropriate
• Exceptional written English, including syntax, register, and nuance
• Care for users, particularly prepaid-plan users who count every dollar

• Flexible working hours
• Competitive salary
• Real opportunity for personal and professional growth