About the team: We are growing the Financial Assistant team at Plata, building intelligent systems that support users in managing their finances, understanding spending, and interacting with financial products in a simple and intuitive way. This team plays a key role in improving customer experience, engagement, and the overall value of our core products. This is a customer-facing product in a regulated environment — accuracy, safety, and trust are non-negotiable.
We use AWS, Go, Python and cloud-based models, yet still flexible to mix ready-made tools and ship our custom solutions. We're all about building systems that deliver real, valuable results for our organization.
You will be a key specialist in a cross-functional team, working closely with backend, mobile and other LLM engineers.
Challenges that await you:
- Work with agentic architectures where needed: tool use, multi-step reasoning, orchestration, and failure recovery
- Design and maintain eval infrastructure (offline test suites, golden datasets, regression harnesses) that gives a reliable signal after every change
- Work with advanced prompts, set up RAG-oriented data sources for efficient retrieval, and evaluate model outputs to ensure accuracy, relevance, and quality
- Implement safety and guardrails: hallucination detection, refusal strategies, factual grounding
- Run online and offline experiments ( canary, A/B) to validate that changes improve real user outcomes
- Own observability from day one: trace LLM calls, monitor quality drift, latency, and cost per session in production
- Collaborate with backend engineers and product to ship reliable improvements, review code, and maintain production health
What makes you a great fit:
- Strong classical ML fundamentals, you understand what's happening inside models, not just how to call their APIs
- Strong eval mindset: you design the measurement system before writing the first prompt, and treat evals as a first-class engineering artifact
- Hands-on experience with search and retrieval: dense/sparse/hybrid, reranking, query understanding — using managed tooling effectively, not necessarily from scratch
- Practical experience with agentic architectures: tool use, orchestration, failure recovery
- Safety-first thinking: guardrails, content policies, graceful degradation under uncertainty, especially in a context where wrong answers have real consequences
- Production ML ownership: observability, latency budgets, cost tracking, regression detection
- Expertise in Python and its ecosystem, as well as language- and framework-agnostic mindset to achieve project goals
- Familiarity with open-source tools to build and evaluate RAG applications.
- Excellent communication skills and ability to explain complex technical concepts to a broad audience of stakeholders
- B1 or higher English level for effective communication with an international team
Your bonus skills:
- Previous experience delivering business-critical ML/AI-powered applications for customer-facing products
- Fine-tuning or distillation of LLMs for production use cases
- Experience deploying open-source models (Llama, Mistral) in private/on-premise environments
- Experience in fintech, banking, or regulated domains