Middle Data Scientist/Machine Learning Engineer in ATB-market

Welcome to join our Middle Data Scientist/Machine Learning Engineer team

Responsibilities: Develop and scale ML solutions for computer vision, personalization and demand forecasting, with tangible impact on P&L and operational metrics networks.

Areas of responsibility: 

• Formulate tasks together with the product, determine success metrics and expected business effect; 
• Design approaches (CV/Recsys/TS), choose metrics (NDCG@K, wMAPE, etc.) and validation methodologies; 
• Full ML cycle: data preparation, experiments, validation, production deployment; 
• Interpretability, drift/degradation control, risk management; 
• Data/model quality support, documentation, hand-off for related teams; 
• Retail cases: CV in the hall/at SCO (queues, anomalies, price tags/planograms, shortages), mission personalization (cold-start, multimodal features), demand-forecast with promo-uplift and cannibalization.

Success criteria (examples):

1. reduction of stock-out by 10%, 
2. 3-4% increase in NDCG@K,
3. 10%-15% reduction in MAPE by SKU/store.

Requirements:

• 3+ years in ML/DS, production cases in CV or testimonials or time series; 
• Advanced Python, classic ML/DL practices, production-level SQL; 
• Understanding on-prem ML lifecycle and interaction with Data Engineering and MLOps.

Would be a plus:

• Retail/FMCG experience; working with videos/images from stores, transactions, promos; 
• A/B-experiments, incrementality, reporting business results.
• Inventory-aware recommendations, anti-fraud on SCO, multimodal features.

Technical stack (on-prem): 

1. Core ML/DL: Python, PyTorch, OpenCV, Albumentations, YOLOv8–v10 or Detectron2, TrOCR or Tesseract. Recommenders: NVIDIA Merlin or Transformers4Rec, implicit (ALS), LightFM, reordering. Vectors/search: FAISS, PostgreSQL+pgvector, Milvus or Qdrant. TFT.
2. Data processing: Pandas or Polars, Spark or Flink.
3. Experiment tracking/registry: MLflow (Tracking/Registry/Serving).
4. Feature store: Feast (self-host).
5. Infra/containers: Docker, Kubernetes or OpenShift.

NLP/LLM-platform 

1. NLP core: Hugging Face Transformers, Datasets, Tokenizers, SentencePiece, spaCy|Stanza (uk), Sacremoses.  
2. LLM serving: vLLM|Hugging Face TGI; TensorRT-LLM|llama.cpp/gguf (by resource profile).  
3. RAG: OpenSearch (BM25) plus re-ranker, chunking and ingest, hybrid search with FAISS|pgvector or Milvus|Qdrant.  
4. Evaluation: ROUGE, BLEU, METEOR, BERTScore, MTEB, Recall@K, MRR, NDCG. Security/PII: Microsoft Presidio. 

The company offers:

• remote or hybrid work format;
• employment on the terms of a gig contract or in the state (reservation is possible);
• paid annual vacation of 24 calendar days, paid sick leave;
• regular payment of wages without delays and in the stipulated volumes, regular revision of wages;
• opportunity for professional and career growth;
• training courses.

Contact Person: Kateryna, phone.0984567857 (t.me/KaterynaB_HR)