Machine Learning Ops2021 – 2022ML Engineer
NYC Airbnb Rental Price Pipeline
MLflow + Weights & Biases experiment tracking
End-to-end rental-price prediction using scikit-learn, MLflow, and Weights & Biases. Estimates typical price from similar properties. Emphasis on experiment tracking, pipeline artefacts, and inference-pipeline deployment rather than EDA theatre.
The Problem & Engineering Constraint
The Core Challenge
Rental models are easy to overfit in a notebook and hard to reproduce once hyperparameters, datasets, and charts scatter across machines.
Technical Architecture & Approach
Engineering Solution & Implementation
Orchestrated pipeline components in MLflow (including random-forest artefacts), tracked datasets, charts, notebooks, and hyperparameters in W&B, isolated dependencies with Conda, configured components with Hydra, and validated the cleaned dataset with Pytest.
Measured Production Impact
Verified Outcomes & Deliverables
Reproducible rental-price inference pipeline with artefact tracking.
Component-level configuration via Hydra and Conda.
Pytest validation of the cleaned dataset before training.
Technologies & Components
System Tooling & Technologies
MLflowWeights & BiasesScikit-LearnHydraCondaPytest