Data Scientist

A. ETA & Trip Intelligence

• Help develop and improve ETA models across contexts (pickup, en-route, multi-stop): feature engineering, model training, offline evaluation and error analysis.

• Perform feature engineering from map and trip signals: road class, intersection density, turn types, time-of-day speed profiles, congestion signals, driver behavior patterns.

• Build and maintain the offline evaluation pipeline: tracking metrics (MAE, p90/p95 error, calibration), comparing model versions and detecting regressions.

• Support GPS and trajectory data normalization: denoising, outlier detection, sampling normalization — providing clean inputs for ETA, traffic and routing models.

B. Traffic Forecasting & Map Signals

• Build and improve spatiotemporal traffic models at road-segment level: speed/flow estimation, incident impact, time-of-day and day-of-week seasonality.

• Generate map-intelligence signals: road-closure detection, abnormal slowdown, and freshness and coverage-quality metrics.

• Analyze and validate traffic data: consistency checks, noise-source detection and proposing data-pipeline improvements with Data Engineering.

C. Dynamic Routing & Decisioning

• Support developing dynamic routing logic (reroute on congestion/incident): implement and test heuristic and model-based approaches. Build and maintain routing evaluation metrics: reroute rate, route deviation, on-time pickup/dropoff, stability and consistency.

• Analyze A/B test results for routing experiments and summarize insights for the Lead DS and Product. D. OCR & Data Understanding

• Help build the OCR and document/image understanding pipeline for field-collected mapping data: signage, house numbers, POI storefronts.

• Perform post-OCR normalization and entity extraction: address, POI name, opening hours; assess confidence and handle edge cases.

• Support building active-learning / human-in-the-loop workflows: selecting samples to label, assessing labeling quality and continuously improving models.

E. Automated Data QA & Approval

• Build models/rules to automatically approve/reject/flag map, POI and address submissions: duplicate detection, spam/fraud detection, consistency checks.

• Design and tune confidence scoring and thresholds for approval policies; evaluate the trade-off between auto-approval rate and false positives.

• Monitor and report auto-approval impact: reduction in manual review, and precision/recall by submission type.

• At least 3 years of Applied Data Science / Machine Learning experience in production. Bachelor's degree (good grade or above) in Applied Mathematics, IT, Statistics or equivalent (top universities preferred: HUST, University of Engineering and Technology, University of Science, National Economics University).

• Proficient in Python for data science: pandas, numpy, scikit-learn; experience with at least one deep-learning framework (PyTorch or TensorFlow).

• Solid statistics and probability foundation: understanding of bias/variance trade-off, calibration, hypothesis testing, A/B testing.

• Experience building and evaluating ML models in production: data-leakage prevention, reproducibility, model versioning.

• Strong data-analysis and cleaning skills: handling missing values, outliers, schema inconsistency; writing SQL and working with large datasets.

• Metric-driven mindset: identifying the right evaluation metrics, detecting regressions and proposing measurable improvements.

• Good English communication (reading/writing): reading technical documents, writing documentation and working with international teams.

• Preferred

• Experience with geospatial data: GPS noise, map-matching concepts, spatial indexing (S2/H3/Geohash), spatial joins.

• Experience building spatiotemporal models: time-series forecasting, sequence models (LSTM/Transformer), graph-based features.

• Experience with OCR/Document AI: text detection, recognition, multilingual, post-processing and entity extraction.

• Basic MLOps experience: MLflow or equivalent for experiment tracking, model registry and monitoring.

• Experience processing big data with Spark/PySpark or streaming with Kafka/Kinesis. Experience with active learning or human-in-the-loop annotation workflows.

• Experience with NLP/text normalization for addresses, place names or accented Vietnamese.