backend developer for financial crime compliance

HireHi; United States; Remote; USD 150,000 - 190,000; Full time

Job summary

Qonto ищет инженера AI/ML для команды AFC, ответственно внедряющего агентные решения в регулируемой среде. Вы будете разрабатывать производственные AI-системы и связанные сервисы, участвовать в архитектуре и обеспечении качества вывода.

Требуется опыт вывода AI-агентов в продакшн и способность принимать обоснованные технические решения без PM, работать в рамках AFC-режимов и стремиться к повышению производительности.

Qualifications

• Опыт вывода в продакшн AI-агентов или рабочих процессов, используемых реальными пользователями, включая оркестрацию, использование инструментов, состояния, структурированные выводы, оценку, повторные попытки и режимы сбоя.

Responsibilities

• Разрабатывать и внедрять производственные AI-системы, агентные инструменты и рабочие процессы.
• Вести проекты от идеи до эксплуатации: сбор требований, дизайн архитектуры, реализация, запуск, поддержка и улучшение.
• Проектировать надежные backend-сервисы, API, базы данных и интеграции с моделями AI.
• Определять оценочные подходы, мониторинг и механизмы проверки результатов; следить за качеством и эффективностью.

Skills

AI agent experience

System design

Independent decision-making

Python familiarity

Job description

Описание

Qonto provides a finance workspace for SMEs, with banking at its core and additional financial tools. Its AI Compliance Tooling team builds AI-powered automation and intelligence tools that help human investigators handle financial crime compliance workflows.

Задачи

Build and ship production AI systems, agentic tools, and AI-powered workflows that gather context from internal systems, orchestrate models and tools, parse structured outputs, and safely handle uncertainty and failure Own projects end to end, from discovery with AFC stakeholders and solution design to architectural decisions, implementation, launch, operation, maintenance, and improvement Build reliable, maintainable, and extensible backend services, APIs, databases, and integrations for evolving AI models and tooling Define evaluation approaches, observability, fallbacks, and human-review mechanisms; monitor output quality, acceptance and edit rates, throughput, and operational impact Reduce investigation lead time and expand automation across AFC workflows through pragmatic, incremental delivery Lead design discussions, anticipate risks, balance speed with quality, and help raise the team’s capability in production agentic systems

Требования

• Experience shipping an AI agent or agentic workflow used by real users, including orchestration, tool use, state, structured outputs, evaluation, retries, and failure modes
• Strong system-design fundamentals across architecture, APIs, databases, integrations, reliability, observability, maintainability, and scalability
• Ability to make independent decisions, communicate trade-offs clearly, and balance speed with quality
• Ability to turn ambiguous operational needs into valuable solutions, challenge assumptions, prioritise scope, and define success without relying on a Product Manager
• Willingness to discover, build, ship, operate, maintain, and continuously improve systems
• Curiosity about AFC and regulated workflows, with the ability to ramp up quickly in a complex domain
• Будет плюсом: prior fintech or compliance experience, Python experience, familiarity with current model providers or agent frameworks

Условия

The team has no dedicated Product Manager; engineers work directly with AFC stakeholders and own solutions from ambiguous needs through measurable production outcomes The team uses a daily 15-minute blocker sync, bi-weekly 1:1s, and lightweight tracking Engineers collaborate directly with operational teams and measure success through investigation lead-time reduction, output quality, human acceptance and edit rates, adoption, throughput, and resources saved The role involves building safe, scalable AI automation in a regulated environment where reliability and auditability matter The team does not build fraud-detection engines or KYC/KYB rule engines; it consumes upstream signals and builds the AI-powered automation and intelligence layer used by human investigators The current system gathers information from 10-15 internal tools and produces a structured compliance analysis for a person to approve or edit, reducing treatment time from approximately 15-20 minutes to about 30 seconds Hiring is open to candidates regardless of which hiring-checklist boxes they tick The hiring process lasts an average of 20 working days