Team Lead / Principal ML Engineer

Travelpayouts is an affiliate adtech platform from the creators of Aviasales. We connect travel bloggers and webmasters worldwide with the largest travel brands: Booking.com, Tripadvisor, Viator, and others.

Why This Role Exists

We have working ML models and accumulated expertise. Now, we need to move from experiments to scale: put models on autopilot, launch new initiatives, and build infrastructure that doesn't require manual intervention every time.

Currently, you will have 2 engineers reporting directly to you. In 3 months, once you settle in and establish processes, the role will grow into Team Lead with team expansion.

What You'll Do

First 90 Days

Here’s what’s realistically in the queue:

• Intent Prediction: We have a hypothesis that we can predict user intent based on behavioral signals — and use this to improve monetization. Models exist in experiments, but not in production. We need to bring them to production.
• Content Classification: The model works but requires manual retraining. We need to automate the cycle: data → training → validation → deployment.
• LLM Gateway: Several product streams want to use LLM capabilities, but each is building its own solution. We need to decide if a unified internal layer is necessary and, if so, design it.

Code and Build:

• Write production-ready code, retrain transformers, assemble eval sets.
• Build MLOps infrastructure: retraining pipelines, feature store, model quality monitoring.
• Participate in Tech Design Reviews and define architectural decisions — be responsible for the stability, fault tolerance, and capacity of ML services.

Connect ML and Product:

• Translate model metrics (F1, ROC-AUC) into business language: conversion, revenue, Time-to-Market.
• Defend technical decisions to the CPO with strong arguments, think about the financial implications of architectural choices.
• Engage in Discovery at early stages — propose MVPs, provide honest effort estimations.
• Explain why an eval set is more important than a new feature — and be able to prove it with data.

Lead the Team (from the 4th month):

• Develop engineers: T-shaped competencies, code reviews, architectural solution discussions, personal engineering example.
• Participate in hiring the next team members.

What We're Looking For

Essential:

• 5+ years in ML/DS — you have deployed models to production and know what that means in practice.
• Proficient Python: OOP, typing, tests.
• Classic ML: gradient boosting, validation, understanding when a neural network is overkill and when gradient boosting is sufficient.
• Experience with text neural network models: BERT-like transformers, LLM APIs, prompt engineering, eval sets.
• Practical MLOps experience: data pipelines, orchestration (Airflow, Kubeflow, or similar), experiment tracking.
• Infrastructure background: Docker, CI/CD, Kubernetes, logging, and monitoring — you understand how high-load services work, not just the models within them.
• Ability to think in P&L terms: translate model metrics (F1, ROC-AUC) into conversions and revenue, defend technical decisions to the business with financial arguments.
• You already have mentorship or leadership experience — and you enjoy it.

A Significant Plus:

• Understanding of Go or experience with Go services: our backend is built in Go, and without this context, some architectural decisions may not be fully clear.
• ClickHouse, Kafka, PostgreSQL.
• Highload experience: designing systems for millions of requests.
• Experience migrating or removing services from a monolith — you understand the cost of architectural decisions at scale.

How We Work

• Anywhere in the world: we are not tied to a location, we pay in dollars, and we love to travel;
• No bureaucracy: convenient, healthy processes, horizontal and open communication, quick idea discussion, and decision-making;
• Compensation: voluntary medical insurance, psychotherapy or foreign language courses, sports activities, and sick leave.