Title : Data Scientist – LLM & GIS Systems
Location : Roseville, CA (Hybrid or Remote)
Type : Contract-to-Hire or Full-time
About LowPropTax
LowPropTax helps homeowners reduce property taxes using data-driven insights and automated appeals. We combine property data, geospatial analytics, and AI to identify over-assessed parcels and build automation pipelines for large-scale appeal filings.
Role Overview
You will design and build LowPropTax’s custom LLM and GIS intelligence stack from the ground up. This includes developing a proprietary language model for property data insights, creating geospatial overlays to visualize property inequities, and automating data workflows across counties.
Key Responsibilities
- Architect and train a custom LLM model using internal tax, assessor, and property datasets.
- Build and fine-tune inference pipelines for automated valuation, appeal reasoning, and evidence generation.
- Develop GIS-based mapping overlays integrating assessor parcels, zoning, and demographic data.
- Automate data ingestion pipelines from public APIs, CSVs, and assessor databases.
- Collaborate with engineering to integrate model outputs into production systems.
- Perform EDA and feature engineering on large, messy, cross-county property datasets.
- Evaluate LLM models for accuracy, explainability, and auditability in compliance contexts.
Requirements
3+ years in data science, AI, or applied ML.Deep understanding of LLMs, embeddings, and transformer architectures (not wrappers like GPT APIs).Experience with PyTorch, Hugging Face, LangChain (core only), and vector databases.Proficiency in Python, SQL, and geospatial tools (GeoPandas, Shapely, PostGIS, Mapbox).Experience with data pipelines and MLOps (Airflow, Prefect, MLflow).Comfort working with county property, parcel, or tax datasets is a plus.Strong problem-solving and autonomy.Nice to Have
Prior experience with public records, real estate analytics, or valuation models.Experience in cloud environments (AWS, GCP) with scalable model training setups.Familiarity with OCR pipelines for document parsing.Why Join
You’ll help shape the intelligence core of a fast-growing proptech startup rooted in real impact — saving thousands of homeowners real money through automation and AI precision.