Ramprasad Rakhonde

Hello. I'm Rampar Sadra Kone. I am from Apollo, Maharashtra. I was born and brought up in Akola. I attended IIT Madras for college. During my college tenure, I was a three-time national analyst for the deep learning competition organized by Amazon and ABN Web. So currently, I'm working with InfoEdge for Nokli.com in Noida. I am a data scientist here, handling search. For seeker search, we have to maintain the complete user seeker search journey. And in that, there are three components. 1st is, like, whenever a user comes and he searches for a query, we have to fetch all the relevant jobs. Then we have to re-rank those jobs. And then there is a personalization layer of re-ranking the last 20 or 40 jobs. So I've developed an innovative algorithm for this last layer, which has given us massive gains of 7.5% in CTR and 7% in applies.

I'd be happy to help you with your concerns. Can you tell me more about what's on your mind?

So by the dataset that we are working with.

I'd like to clarify my role as a grammar editor for interview transcripts. My main goal is to correct grammatical errors, remove filler words, and maintain the speaker's natural voice and structure.

It depends on what NLP based system we are talking about. If it's the one that can be benefited from embeddings, then the embeddings generated by LLM can be used. If it's the one that basically uses algorithms that can be used directly as well.

I'd be happy to help you review the transcript. Please go ahead and provide the text.

It is trying to fetch 10 values from points, which are basically name, location, and category. Their category should be hotel or restaurant, and location shouldn't be null. And then we are trying to order them by length of name. But the problem is over here that the limit 10 and the first statement are separated by a semicolon. So that means the first will run entirely and then the second will run. So that is not actually achieving the purpose. So, after this, there should not be a symbol.

So in this Python code, we are trying to read a data frame. If the file is not found, then they're using the error. But what's what if there is some file that is found, but there is another error, then that error is not being caught. So I believe there will be a difficulty, and there is a bug. So instead of file not found, we can directly put try and last exception block over there.

So we'll talk about this in several steps. 1st of all, data gathering, then doing the EDA on the data, trying to get the feel of the data. How is the data? What are the features? Then we will go to the feature selection methods. There are feature selections and preprocessing of the features, then the model decision that is to be made. Then what is our metric or here it is, we can have high accuracy metric. And then we will have to go into the actual model building part. So let's start with the data collection. So for the data collection process, if we already have data with us, then that's great. Then we will face that data, and we will perform EDA on it. If there are any medium findings, what about other findings? Those findings will be helpful for the further process. So if there are any correlated features, we will get to know that as well. And in the feature selection method, we will check for correlation between the features. What if there are many features to consider, then we can check which features should be used. In the middle of preprocessing, if it is categorical data, then we will convert it to the correct form. If it is new categorical data, we will convert it to numbers by using any encoding method depending on the algorithm we are using. For here, I'll just assume that we will be using random forest, and it will suffice for the given problem. We will encode the categorical data, and numerical data will be standardized or normalized. And then that data can be used again to build a smaller model. Then coming to the matrix part, we have to create a high accuracy matrix. It works most of the way, and that should match. If it was a precision or recall oriented metric, then we have to focus on one part where we should maximize precision, we should maximize recall, having a decent value in the other field. Then after this, it comes down to validating the model and the validation dataset and for datasets similar to the production dataset. Then after that, the deployment part will come.

I'm not sure about that either.

So as I mentioned previously, the best approach would be to use the embedding generated by instead of the previous embedding that you are using in NLP based processing. That will be the fastest approach and give us a lot of gains. The other way can be using the generative features of LLM in the end processing. So if there is any other task, then basically we can use LLM or address it for a bit more clarity and all. This is a basic example I'm giving. It's a small task, and I would ask you to consider LLM. But depending on what we are doing, LLMs can be integrated into their processing as well.