Retello is India’s first AI-powered shopping assistant. Instead of browsing spec sheets and wading through review noise, you ask a natural-language question — “best gaming phone under ₹20k” or “best camera phone under ₹15k” — and get a ranked, reasoned answer tailored to that use case. I built two of the core systems that make this possible: the scraping infrastructure layer and the benchmarking engine.
Scraping infrastructure
Retello pulls product data from Amazon, Flipkart, Croma, Jio, and other major Indian e-commerce platforms. I built and own the entire infrastructure layer: auto-scaling scraper fleets, dynamic server provisioning, orchestration of modules across sources, and the pipelines that turn raw crawl output into usable product records.
For each product, the pipeline goes beyond basic listing data. It aggregates reviews from across the internet, extracts granular specifications, and processes everything through AI pipelines to produce structured, normalized records. The individual site-specific scrapers were built by others on the team; I designed and managed the infrastructure that runs, scales, and coordinates them.
Benchmarking engine
The benchmarking engine is what lets Retello answer ranked, use-case-specific queries rather than just returning raw search results. I built it end to end.
The core is a comparison system that evaluates products at three levels:
- Component-level rankings — individual spec comparisons (camera sensor, battery, processor)
- Product-level rankings — aggregate scores per device
- Compound rankings — use-case-weighted scores (gaming, photography, battery life, value)
A query like “best phone for photography under ₹20k” resolves by pulling the relevant product set, weighting specs against the photography use case, and surfacing the ranked result. The entire pipeline is automated — no manual curation, no hand-tuned lists.
What it ships
Retello launched in early access with a waitlist. The first 1,000 users get exclusive access codes.