AI
Engineering
Research
Validated 10,000+ property records into a land-value forecasting tool
LandPulse · University of Westminster · Oct 2023 – Mar 2024
Context
Land prices in the Colombo district are opaque and negotiated informally, leaving buyers without a reference point.
Problem
Buyers and sellers needed an estimate of current value and likely growth, from messy public data.
My role
Full-stack developer in a university team project, owning the data pipeline and the product surface.
Constraints
- Public property data was inconsistent and incomplete
- Academic timeline of one semester
- Team of student developers with mixed availability
What I did
My contribution
- Supporting work: built RESTful APIs and React components for search and valuation flows
- Supporting work: built MongoDB data pipelines and cleaned/validated 10,000+ property records
- Supporting work: integrated the growth-forecasting model output into the product UI
Owned by the wider team
- Model selection and training were shared with two teammates
- Report and evaluation were team-authored
What I'd do differently
- I would show a confidence range rather than a single predicted price — a point estimate over-promises on noisy data.