ML Infrastructure Engineer (Python)

About the role

Responsibilities

• Technical Leadership & Ownership: Take full ownership of the end-to-end Decision-Making System (DMS) lifecycle, including system architecture, deployment pipelines, production monitoring, continuous code cleanup, and performance optimization.
• Cross-Functional Collaboration: Partner closely with Risk Analysts and Data Scientists to translate complex business logic, risk rules, and ML scoring models (application, behavioral, fraud) into production-ready software.
• Team Leadership & Delivery: Implement core logic of DMS system to suit the needs of InDrive's lending vertical. Code reviews, release management, and performance evaluations while fostering an efficient, agile environment.
• Process Optimization: Reduce time-to-market for credit strategy updates by modernizing delivery workflows, introducing automated regression/retrospective testing, and eliminating dependencies on legacy release cycles.
• Architecture & System Integration: Drive the evolution of our high-volume microservices ecosystem (Golang, AWS). Integrate internal and external services (credit bureaus, collection platforms, fraud detection services) via REST APIs and event-driven pipelines.

Qualifications

• Fintech & Credit Decisioning Expertise: 7+ years of total backend engineering experience, with deep domain expertise in financial technology, credit conveyors, loan processing, debt collection systems, or automated decisioning engines.
• Risk-Ops experience: be familiar with any popular decision-engine (gorules, n8n, camunda, SAS RTDM, FICO Blaze Advisor etc…)
• Stack: Strong expertise in Golang / Python, and build tools (Gitlab CI), AWS-related experience will be a huge plus.
• Data Management: Proficiency with relational and non-relational databases (MySQL, AuroraDB, DynamoDB, GBQ).
• DevOps, CI/CD & Observability: Hands-on experience with modern CI/CD tools (GitLab CI, GitHub Actions), containerization (Docker), and system monitoring tools (Prometheus, Grafana).

Benefits

• Help us challenge injustice by creating fair choices for millions of people across 47 countries.
• Develop your professional skills with access to mentoring, career consulting, and learning programs.
• Collaborate with teams around the world and gain international experience through our Global Talent Exchange Program.
• Engage in company-wide challenges, awards, sports activities, employee-led social impact and volunteering projects.
• Work alongside people who take initiative, speak openly, and challenge themselves to grow.
• Improve your language skills through co-financed courses and internal speaking clubs.
• Annual salary from 54 200 EUR gross.

Ищем ML-инженера, который умеет не просто обучать модели, а доводить их до production: от данных и feature engineering до деплоя, мониторинга и улучшения качества в реальных системах.

Что будет в работе

• Разработка и развитие ML-моделей: классические ML (классификация, регрессия, ранжирование), deep learning, NLP.

• Проектирование и реализация end-to-end ML-пайплайнов: от данных и feature engineering до обучения, валидации и деплоя.

• Работа с данными: ETL/ELT, подготовка датасетов, feature store, работа с большими объёмами данных.

• Деплой моделей в production: Docker, Kubernetes, REST/gRPC API, batch/online inference.

• Настройка MLOps: MLflow/W&B, версионирование моделей и данных, CI/CD/CT, автоматизация переобучения.

• Мониторинг моделей в production: метрики качества, latency, throughput, data/model drift, алертинг.

• Участие в архитектурных решениях: выбор подходов, оценка trade-off’ов, влияние на стабильность и масштабируемость системы.

• Взаимодействие с data scientists, инженерами, аналитиками и бизнес-стейкхолдерами.

Что важно

• Коммерческий опыт в ML/ML Engineering от 3 лет, уровень Middle+ / Senior.

• Уверенное владение Python:pandas, NumPy, scikit-learn, PyTorch или TensorFlow.

• Опыт построения и деплоя ML-моделей в production (не только Jupyter/прототипы).

• Понимание классического ML: supervised learning, feature engineering, валидация, метрики качества.

• Опыт работы с данными: SQL, ETL/ELT, Spark/Databricks (желательно), работа с большими датасетами.

• Опыт работы с Docker, базовое понимание Kubernetes, CI/CD (GitLab CI, GitHub Actions).

• Опыт настройки MLOps: MLflow/W&B, версионирование моделей/данных, автоматизация пайплайнов (Airflow, Kubeflow, Databricks Workflows).

• Понимание мониторинга моделей в production: метрики, drift detection, алертинг, retraining.

• Умение работать автономно, брать задачу целиком и доводить до результата.

• Умение переводить бизнес-требования в технические решения и аргументировать свой выбор.

Будет плюсом

• Опыт работы с LLM/GenAI: RAG, embeddings, vector databases (FAISS, Milvus, pgvector), prompt engineering, fine-tuning (LoRA)

• Практика LLMOps: MLflow, Langfuse, LangSmith, Weights & Biases, evaluation harnesses (Ragas, DeepEval)

• Опыт работы с agentic systems: LangGraph, CrewAI, multi-agent workflows

• Опыт работы с облачными ML-платформами: AWS SageMaker, Azure ML, Vertex AI

• Опыт работы с feature stores (Feast, Tecton), model registries, A/B testing, canary deployments

• Знание паттернов MLOps: CI/CD/CT, drift detection, rollback, monitoring (Prometheus, Grafana, Evidently, Arize)

• Английский язык для работы с зарубежной документацией, командами и стейкхолдерами.

Английский

• Для проектов без английского язык не требуется.

• Для европейских и американских проектов нужен английский от B2+.

Процесс

• Короткое общение с рекрутером.

• Техническое интервью / TechCertify-скрининг.

• Передача на подходящие проекты клиентам.

Если вам близок инженерный подход к ML, нравится не просто «обучать модели», а доводить их до production и влиять на реальные продукты - будем рады пообщаться.

The role

Nebius builds the infrastructure serious AI teams run on — GPU clusters, inference runtimes, agent development environments, data pipelines — all of it purpose-built for the most demanding AI workloads. What we are now building is the ecosystem function that ensures the best AI companies choose to build on us, integrate with us, and stay. 

As a Forward Deployed Engineer, Ecosystem, you will sit at the intersection of solution architecture and hands-on engineering. You assess how partner products actually work on our stack, define the reference architecture for each integration, build the working prototype that proves it, and translate what you find into product requirements that shape what Nebius ships next. 

Your responsibilities will include:

Solutioning & Architecture 

• Design and prototype integrations between partner products and the Nebius platform — fast, hands-on, and technically sound
• Define reference architectures for partner integrations — not just what works, but how it should work at scale and in production
• Scope partner architectures against our platform — how does this product actually work on our stack, where does it snap together, where does it break
• Build production-quality proof-of-concepts across the AI stack including agentic pipelines, RAG architectures, inference optimization patterns, and multi-model orchestration
• Produce working proof-of-concepts that serve as the starting point for product creation — not a requirements doc, a working thing
• Maintain a library of reference architectures and integration patterns that internal product and engineering teams can build from 

Technical Partner Scoping 

• Work directly with partner engineering teams to scope, prototype, and progress integrations
• Assess partner architectures honestly — if the integration is painful, that is signal; if it snaps together in a weekend, that is also signal; report both
• Provide technical guidance to partners on how to maximize performance, reliability, and cost efficiency on Nebius infrastructure
• Produce technical scoping that gives your pod partner and internal teams a clear picture of integration feasibility, depth, and complexity 

Internal 

• Translate external integration findings into actionable product requirements for Nebius platform teams
• Work with ISV partners, SI teams, and field teams to scale solution adoption and drive revenue once a solution is ready
• Surface recurring architectural patterns and integration gaps to inform platform roadmap decisions
• Participate in platform planning as the technical voice of what you are seeing and building in the field 

Ecosystem Presence 

• Represent Nebius at hackathons, in open source communities, and at technical events
• Build in public — demos, reference architectures, and integrations that establish Nebius as the platform serious AI builders choose
• Stay current with the AI tooling ecosystem — you know what shipped last week and what it means for our stack 

Platform focus areas: 

Depending on your background and mutual fit, you will focus on one or more of the following: 

• Agentic — agent frameworks, memory systems, tool integration, orchestration, MCP, guardrails
• Managed Inference — inference runtimes, model serving, optimization tooling, speculative decoding, KV-cache routing
• IaaS / Managed Infrastructure — cloud-native integrations, GPU orchestration, enterprise platform connectors
• Data — vector databases, retrieval systems, RAG architectures, data pipeline integrations, synthetic data tooling 

We expect you to have:

• 6+ years of hands-on engineering experience in AI application development, ML systems, or AI infrastructure
• Deep working knowledge of the AI developer stack — LLM APIs, inference runtimes, orchestration frameworks, vector databases, RAG architectures, agentic pipelines — built through shipping, not reading
• Hands-on experience with agentic frameworks such as LangChain, LangGraph, CrewAI, AutoGen, or equivalent
• Strong Python programming skills and comfort prototyping end-to-end AI systems quickly
• Experience defining reference architectures and technical patterns — not just implementing them
• Proven ability to move from idea to working prototype fast — you have shipped meaningful things under time pressure and found it energizing
• Experience building integrations across APIs and developer platforms — you understand where the complexity actually lives
• Comfortable working across both external partner engineering teams and internal Nebius product and engineering teams simultaneously
• Strong technical communication — you can explain architecture decisions and integration findings to a founding CTO and a non-technical partner lead in the same day 

It will be an added bonus if you have: 

• Experience with inference frameworks and optimization: vLLM, SGLang, TensorRT-LLM, speculative decoding, quantization, batching, KV-cache routing
• Familiarity with NVIDIA's software stack: CUDA, TensorRT, NeMo, or equivalent
• Experience with multimodal AI models — vision-language, speech, or structured data
• Won or placed at major AI hackathons in the past 12 months
• Worked as a developer advocate, solutions engineer, or technical partner manager at a leading AI platform or developer tooling company
• Been an early engineer at a YC-backed AI startup — you built the product under real constraints
• Open source projects or public demos with meaningful community adoption
• Proficiency with DevOps tools: Docker, Kubernetes, Git 

Preferred technical stack: 

• Languages — Python
• ML frameworks — vLLM, SGLang, TensorRT-LLM, Transformers, OpenAI / Anthropic SDKs
• Agentic frameworks — LangChain, LangGraph, CrewAI, AutoGen, smolagents, or equivalent
• Vector databases — Qdrant, Weaviate, Milvus, pgvector
• API and web frameworks — FastAPI, Flask
• DevOps — Kubernetes, Docker, Git
• Cloud platforms — AWS, GCP, Azure 

Key Employee Benefits:

• Health Insurance: 100% company-paid medical, dental, and vision coverage for employees and families.
• 401(k) Plan: Up to 4% company match with immediate vesting.
• Parental Leave: 20 weeks paid for primary caregivers, 12 weeks for secondary caregivers.
• Remote Work Reimbursement: Up to $85/month for mobile and internet.
• Disability & Life Insurance: Company-paid short-term, long-term, and life insurance coverage.