Job Title

Applied Scientist — Property Intelligence (AVM & Knowledge Graph)

Company

Synergy Marketing Technology Limited (building HomeDash, our real-estate platform subsidiary)

Location

Kwun Tong, Kowloon — Suite 2556, Level 25, Two Harbour Square, 180 Wai Yip Street

Employment Type

Full-time · Mid–Senior

Job Highlights

• Own AVM v1→v2 for Hong Kong, expand to new regions
• Build Property Knowledge Graph (estates/buildings/units/aliases/POIs)
• Design Liveability Index; power search, ranking & SEO
• Ship ML to production with real monitoring & impact

About the Role

You’ll build the core intelligence behind HomeDash—clean, connected data and trustworthy valuations that drive search, discovery, and lead generation. This role blends research depth with product pragmatism: you’ll design models, stand up data pipelines, and put them in users’ hands quickly.

Responsibilities

• AVM: data sourcing, feature engineering, model selection, backtesting, calibration (confidence bands), drift monitoring
• Knowledge Graph : schemas & entity resolution (aliases, addresses, geos), listing de-dup , location standardisation
• Liveability Index : define metrics (transport, schools, amenities, green, noise/air, walkability) and ship to UI/ranking
• Estate/Listing Q&A (RAG) : parsing, chunking, retrieval, evaluation & safety guardrails
• MLOps : batch/real-time inference, monitoring, model registry/feature store; partner closely with our GZ engineering team
• Ground truth & evaluation : labelling guidelines, error analysis, offline/online metrics; clear documentation

What Success Looks Like

• 90 days: AVM v1 (districts) with baseline MAPE & confidence bands; KG v1 with >95% alias resolution on top estates/buildings; liveability framework agreed
• 6 months: AVM v2 uplift; liveability in ranking/filters; RAG QA v1 with eval set & guardrails; model/service monitoring live

Requirements

• Advanced degree or equivalent experience in CS/Stats/Applied Math/EE/Geospatial
• 3–6+ years hands-on ML for tabular & geospatial (time-series/entity resolution a plus)
• Proven end-to-end productisation (experimentation → deployment → monitoring)
• Statistical rigour (calibration, backtesting, bias/fairness) and strong communication
• Language: English required; Cantonese preferred; Mandarin a plus

Tech Stack (nice to know)

Python, Pandas/NumPy, PyTorch/TF, XGBoost/LightGBM, scikit-learn, GeoPandas/QGIS, Airflow/Prefect, MLflow, Feast, Postgres/BigQuery, pgvector/FAISS, Elastic/OpenSearch, Docker/K8s, dbt

Nice to Have

Real-estate/fintech/maps, address normalisation/geocoding, mobility/remote sensing, cross-border collaboration experience

Benefits & Work Style

Medical · Learning budget · High ownership, fast shipping culture

Please include a short note on an ML system you shipped (goal → method → result).