Machine Learning Engineer – Forecasting, Recommendation Systems & MLOps

We are looking for a hands-on Machine Learning Engineer who can design, build, deploy and monitor production-ready machine-learning solutions. The selected freelancer should be comfortable working across the full ML lifecycle, including business-requirement analysis, data preparation, feature engineering, model development, evaluation, deployment, monitoring and secure handling of client data. This is not a research-only or notebook-only role. You should be able to convert business problems into reliable ML systems that can be integrated into real applications and operational workflows. Projects may involve forecasting, recommendation systems, classification, regression, clustering, anomaly detection, customer segmentation, natural-language processing, semantic search and retrieval-based systems. Key Responsibilities: - Translate business and client requirements into measurable ML objectives, datasets, baselines, acceptance criteria and delivery plans. - Explore, clean, transform and validate structured, semi-structured and unstructured datasets. - Identify data-quality risks, missing data, leakage, bias, drift and sampling problems. - Build reproducible training, evaluation and inference pipelines. - Develop forecasting solutions for demand, sales, inventory, staffing, capacity, pricing and operational planning. - Build recommendation and ranking systems using collaborative filtering, content-based methods, embeddings or hybrid approaches. - Develop classification, regression, clustering, anomaly-detection, churn, propensity and segmentation models. - Build natural-language applications such as text classification, document extraction, semantic search and retrieval-augmented systems. - Apply feature engineering, transfer learning, fine-tuning and data-augmentation techniques where appropriate. - Select evaluation strategies and metrics that reflect both technical performance and business outcomes. - Perform error analysis, sensitivity testing, ablation studies and comparison against credible baselines. - Package models for batch or real-time inference. - Develop reusable model APIs and backend services. - Deploy models using Docker to cloud, local, private-network or on-premises environments. - Implement monitoring for data quality, latency, failures, drift, resource usage and model performance. - Support model retraining, versioning, rollback and controlled releases. - Work with software and data engineers on APIs, data contracts, schemas, orchestration, testing and production reliability. - Document datasets, experiments, assumptions, limitations, model behaviour, deployment procedures and operational runbooks. - Handle client data securely using access controls, encryption, pseudonymisation, retention controls and auditable processes. The role requires practical ownership across data preparation, modelling, deployment, monitoring and production support. Required Skills and Experience Applicants should have: - Approximately 3–5 years of relevant professional experience in machine learning, applied data science, ML engineering or a related field. - Evidence of deploying at least one model beyond a notebook into a tested and usable production or operational environment. - Strong Python programming skills and the ability to write clean, testable and maintainable code. - Good SQL skills and experience working with relational or analytical databases. - Strong understanding of supervised and unsupervised learning. - Knowledge of regularisation, feature engineering, probability, optimisation and experimental design. - Experience with classical ML libraries such as scikit-learn, XGBoost, LightGBM or CatBoost. - Practical experience with PyTorch, TensorFlow/Keras or another deep-learning framework. - Strong experience in either forecasting or recommendation systems, with working knowledge of the other. - Experience selecting metrics appropriate to the specific business problem. - Familiarity with model serving, APIs, containerisation and production monitoring. - Good software-engineering fundamentals, including Git, code review, testing, packaging and reproducibility. - Understanding of privacy, secure data handling and responsible AI. Relevant Technologies Experience with several of the following is expected: - Python and SQL - scikit-learn - XGBoost, LightGBM or CatBoost - PyTorch or TensorFlow/Keras - Hugging Face Transformers - Pandas or Polars - Spark - FastAPI or Flask - REST APIs - PostgreSQL or similar databases - ARIMA, SARIMA, Prophet or other forecasting approaches - Collaborative filtering, embeddings and ranking models - Airflow, Prefect or Dagster - MLflow or Weights & Biases DVC - Docker - Kubernetes - GitHub Actions or GitLab CI - AWS, Azure or GCP - Parquet and modern data-pipeline formats Additional Relevant Experience The following would be advantageous: - Feature stores - Vector databases - Streaming-data systems - Distributed training - LLM evaluation or retrieval systems - Model quantisation, pruning or distillation - ONNX or TensorRT - A/B testing or causal inference - Work in regulated, privacy-sensitive or multi-client environments Expected Deliverables Depending on the project, deliverables may include: - Cleaned and validated datasets - Feature-engineering pipelines - Reproducible training and evaluation code - Forecasting or recommendation models - Classification, regression or anomaly-detection models - Batch or real-time inference pipelines - Model APIs and backend integrations - Dockerised model services - Cloud or local deployment configuration - Model monitoring and drift-detection components - Automated tests and CI/CD configuration - Model documentation and operational runbooks Application Questions Please answer the following five questions in your proposal: 1. Describe one machine-learning solution you developed and deployed beyond a notebook. What business problem did it solve, what was your specific contribution, and how was it deployed and monitored? 2. Describe your experience with forecasting or recommendation systems. Explain the modelling approach, features, validation strategy and evaluation metrics you used. 3. How do you identify and prevent data leakage, overfitting and unreliable evaluation results? Include an example from a real project where possible. 4. How would you package and deploy a machine-learning model for reliable batch or real-time use? 5. Explain your experience with APIs, Docker, cloud deployment, model versioning and rollback. Share links or examples of relevant ML projects. These may include forecasting, recommendation systems, classification, deep learning, NLP, model deployment or MLOps projects.