We are currently looking for an ML Engineer (Recommendation Systems) to join our Data Platform Development team. You will be responsible for developing recommendation systems and helping us transition from rule-based solutions to ML.
Recommendations currently operate on a rule-based baseline in production; the next step is to move to ML and build a full cycle of experiments around them.
This is a non-standard task, and it's important to understand this before applying. The assortment a user sees is determined by the delivery geozone and product availability with specific local merchants. The catalog updates quickly, products are perishable, and cold start for SKUs is the norm, not an edge case. There is little personal history: gifts are bought only a few times a year. It's a two-sided marketplace — user relevance must be balanced with merchant economics.
Responsibilities
First three months:
- Understand the current baseline and data, identify where quality is actually being lost;
- Design and set up the model training and inference pipeline;
- Formulate a working ML experiment cycle: offline validation → A/B testing on live traffic → decision based on experiment results.
Further, as the system matures:
- Priority #1 for the coming year remains recommendations: ranking quality, personalization, handling cold starts.
- In parallel, the platform part will grow — feature store, experiment framework, model monitoring.
- In the future, there are directions in CV (product classification, image search) and demand forecasting.
Requirements
- Approximately 4+ years of experience in production ML in recommendations, search, or ranking, involving the full development cycle – from model training to production launch;
- Experience in developing modern RecSys models and launching them into production with a measurable impact on product and business metrics;
- A/B tests: you have designed, launched, and analyzed them yourself (not just read the results);
- Autonomy: you can design a system from idea to production without daily supervision.
Technical Skills:
- Python (production-grade code);
- PyTorch or TensorFlow (you have actually trained models, not just used them for inference);
- SQL (complex analytical queries) for data collection and preparation;
- Linux, Git, Docker.
Engineering Mindset:
- You know how to justify when a model is needed and when a heuristic is sufficient – in both directions;
- You discuss trade-offs: latency vs. quality, complexity vs. maintainability;
- You validate results in a way that can be trusted;
- You are ready to read someone else's code in an unfamiliar language and negotiate contracts between teams.
A Big Plus Would Be:
- multimodal embeddings (CLIP/SigLIP-like), working with product content representations;
- Vector databases: Qdrant, Milvus, Pinecone;
- Feature stores: Feast, Tecton;
- MLOps: MLflow, experiment tracking, model registry;
- Causal inference, uplift, interleaving;
- Kubernetes, cloud (Yandex Cloud / AWS);
- Knowledge of the e-commerce domain: marketplaces, two-sided markets, supply constraints.
Conditions
- Employment: employment contract;
- Work schedule: 5/2;
- Remote work format.
What Nice Things We Do for Employees
- We offer VHI with dental coverage or compensation for medical services, as well as a psychological support service;
- We care about physical health — we compensate for expenses on sports activities or the purchase of sports equipment;
- We compensate for the cost of equipment you purchase for work;
- We create conditions for development: we pay for training, conferences, and meetups necessary for work;
- We provide access to a corporate loyalty program: personal discounts on Flowwow and offers from partners.
Our Values
Care and Love
Our product is based on people's desire to please each other and be together. As long as love exists, Flowwow will have users.
Development and Growth
Flowwow is rapidly developing and gladly supports employee initiatives.
Culture of Mutual Assistance
We employ people who are involved in the process and care about what they do.
Openness to Change
Everyone can show initiative, offer suggestions, and quickly see their contribution to the overall success.
Charity
One of Flowwow's goals is to help people. We are united by social responsibility and a willingness to participate in special charity events.
