Indian Used-Car Price Predictor AI model
Budget: ₹1,500 – ₹12,500 INR
I want to stand up a full, production-ready price-prediction service for the Indian second-hand car market. To get there, I need you to handle everything end-to-end—from pulling the raw listings to serving a live model on Google Cloud.
Data gathering
Start by scraping both authorised dealer portals and popular user-listing sites. Make sure the collection process respects each site’s T&C and captures the usual variables buyers care about: make, model, year, kilometres, fuel type, transmission, location, asking price and any other signals you know improve accuracy.
Data preparation
Once the raw CSV/JSON is in hand, clean obvious errors and outliers, normalise numeric fields, and leave categorical values encoded in a way that plays nicely with regression algorithms.
Modelling
Stick with regression analysis (no neural nets needed here). I’m open to linear, polynomial, tree-based or ensemble regressors—whatever scores best on cross-validated MAE/RMSE. Track experiments so hyper-parameter tuning is reproducible.
Deployment
Package the winning model and deploy it as a REST endpoint on Google Cloud (AI Platform Prediction or Cloud Run—whichever you prefer). Return a simple JSON price estimate for a supplied feature dict. Secure the endpoint with an API key.
Expected hand-off
• Cleaned & annotated dataset
• Well-commented Python code/notebooks
• Trained model artefact
• Live GCP endpoint + deployment script
• README or video walkthrough showing how to hit the API
I’m only considering serious, India-based freelancers who have done full-stack ML delivery before. If that sounds like you, tell me roughly how many listings you plan to scrape and which regression technique you’d start testing first.
Data gathering
Start by scraping both authorised dealer portals and popular user-listing sites. Make sure the collection process respects each site’s T&C and captures the usual variables buyers care about: make, model, year, kilometres, fuel type, transmission, location, asking price and any other signals you know improve accuracy.
Data preparation
Once the raw CSV/JSON is in hand, clean obvious errors and outliers, normalise numeric fields, and leave categorical values encoded in a way that plays nicely with regression algorithms.
Modelling
Stick with regression analysis (no neural nets needed here). I’m open to linear, polynomial, tree-based or ensemble regressors—whatever scores best on cross-validated MAE/RMSE. Track experiments so hyper-parameter tuning is reproducible.
Deployment
Package the winning model and deploy it as a REST endpoint on Google Cloud (AI Platform Prediction or Cloud Run—whichever you prefer). Return a simple JSON price estimate for a supplied feature dict. Secure the endpoint with an API key.
Expected hand-off
• Cleaned & annotated dataset
• Well-commented Python code/notebooks
• Trained model artefact
• Live GCP endpoint + deployment script
• README or video walkthrough showing how to hit the API
I’m only considering serious, India-based freelancers who have done full-stack ML delivery before. If that sounds like you, tell me roughly how many listings you plan to scrape and which regression technique you’d start testing first.