HN user

timotta

3 karma
Posts9
Comments14
View on HN

One of the activities involved splitting into groups to imagine what OLX would look like in the future and present it. After each presenttion, we were asked to imagine a post-future and present again. This cycle repeated several times, until we eventually reached a point where AI would automatically identify products you no longer needed, sell them to someone who did, pick them up from your home, deliver them to the buyer, and transfer the money to you, all without interrupting you at all. It would feel as natural as electricity or water service does today, quietly running in the background and charged via automatic debit.

It was a great reminder that, in data science, the smartest thing you can do isn’t always building a new algorithm: It’s helping people make slightly better decisions at the right moment. After all, the Agile Manifesto wasn’t written for data scientists, but maybe it should’ve been: working insights over perfect models, collaboration over documentation, and delivering value over chasing accuracy.

This is the story of how we improved the search experience at OLX Brasil in early 2024 by introducing semantic search. All major players in search are investing in hybrid approaches, because they’re not only proven to deliver better results, but also offers better scalability for product development by avoiding manual tuning of lemmatization, synonyms, and other linguistic rules.

[dead] 2 years ago

This brief introduction about myself is meant to provide a clearer response to the developers who often ask me how to become a hybrid developer like me. The answer is not straightforward, as my experience has been quite erratic. There was no clear learning path. As I’ve explained, most of my learning was unstructured and driven by business needs. However, there are indeed some courses and books that helped me along the way to hybridizing my knowledge. Without further ado, here are the main courses that helped me.

[dead] 2 years ago

Sometimes a traditional feature brings more value than a machine learning model, and sometimes we need a machine learning model to realize this. This text tells the story of a great success, which could have been a great failure if we had clung to the solution instead of the problem.

[dead] 2 years ago

This article explains Podium, a search relevance algorithm created by the OLX Brazil Search team to replace the globally renowned BM25. It resulted in increased clicks, contact intentions, and reduced complaints about search quality, and was launched in April 2021.

[dead] 2 years ago

This is the story of how a simple mitigation solution led us to discover the root cause of a difficult-to-replicate bug, originating from a lack of knowledge about a technology.

[dead] 3 years ago

Another politician threatened to sue me through a Facebook message if I didn’t take down his profile because it would be defamation. The interesting thing is that this politician was highly rated on the app. I explained this to him and removed him from the platform. Then he asked to come back and I refused.

This is the story of how I invented Ajax (Asynchronous JavaScript + XML) despite it already existing, and how a junior developer would greatly benefit from mentorship at the beginning of their career.

[dead] 3 years ago

This is the story of how we managed to prevent “data collect” fraud through a data collection!

TLDR generated by GPT3:

In 2006, developers at Globo.com wanted to break down silos between different sites by allowing developers from other sites to offer GMC registered videos in a simple and standardized way. However, the platform's rigidity made it challenging to display dynamic video content on static HTML pages. To solve this problem, they developed standardized visual widgets on the client-side that obtained information from WebmediaAPI by returning a JavaScript call to a callback function. However, they needed server-side rendering to maintain relevance in organic searches. To achieve this, they rendered JavaScript widgets on the server using Rhino. Despite facing challenges, they successfully developed a client-side widget platform that enabled all websites to offer videos consistently and created a server-side rendered video product that could be indexed by Google.