@Hi_Mrinal Allnighter never works for me wasted whole week to learn new stuff only to find out I understood 30% of the actual context when I was tested
Wrapped up my first data-gen project: automated an end-to-end pipeline to generate synthetic Hindi conversational data, capped at 1k high-quality rows with a QC layer to filter weak/inconsistent samples. Good hands-on reps in prompting & data quality
@ItsEzekielTakam Basically I'm using llm to generate the raw conversational pairs in hindi then running a local Python script to validate the JSON structure nd filter out low quality outputs before saving