On our mission to make FindYogi the most useful platform for making buying decisions we come across problems that are generic to most web applications that deal with a lot of data. There are particularly the following areas of data that we are working to improve. Since capturing and analysing data is the core of our product, we want to be perfect at it.
1. Auto clustering of products from different merchants – While we collect various offer listings from ecommerce companies, the product names mentioned by them vary. It could be a simple variation like “Samsung Galaxy S3” vs. “Samsung Galaxy S III” or various features of product mentioned in the name itself. The current process of tagging a merchant offer with its relevant product on FindYogi is part manual and part automated. Since the tagging has to be 100% accurate, full automation may not be possible. We want to be able to make this as much automated as possible; as an immediate goal, atleast decrease the manual effort to half of what it is now.
2. Feature Score – The algorithm to calculate feature score for each category has to be hard coded currently. In the long run, we want to be able to share a interface where various conditions can be input by the category analyst without involvement of an engineer.
3. Better matching suggestion – The current suggestion of “Best compared with” in the product page needs to be more personalized depending on the filters set by the user in the listing page. Considering more params and yet maintaining the same speed of serving the data is an issue.
4. NLP based search – The autosuggest in our search is pretty inefficient in terms of match and speed. The ideal user expectation for search is to work like Google, in terms of match and speed. That is tough but we can get closer by spending more hours tweaking the current code. What we really want to do is serve natural language queries like “samsung mobiles with 2 GB RAM” or “4g enabled phones under 15K”. And also such queries across categories that we grow.
There are more problems which are even more complex but little too early to talk about in public. Some of these problems are only a matter of time, while others can be too difficult to solve with perfection. If you think any of these interests you and you want to work with us, write to KickSomeAss [at] findyogi [dot] com. We pay well, in cash and ESOPs.
For the reference, we currently use LAMP stack with Codeigniter, Python with Scrappy, R (for Score analysis).