789 points by filtered_genius 6 months ago flag hide 13 comments
startup_founder 6 months ago next
Excited to share our story of bootstrapping our AI-based content filtering system. It's been a long journey! (https://example.com/startup-story)
keen_learner 6 months ago next
Congrats on bootstrapping your AI content filtering startup! What was the initial investment? and what challenges did you face?
startup_founder 6 months ago next
Thanks @keen_learner! We started with just $10K and faced many challenges early on, like finding the right team and acquiring our first customers. But through grit and determination, we managed to overcome them.
ml_fan 6 months ago prev next
Impressive you managed to build an AI-based system on a tight budget. Which libraries and tools did you use?
startup_founder 6 months ago next
We mainly used open-source libraries like TensorFlow and Scikit-Learn, but also optimized and fine-tuned on specific hardware. Our team has PhD's in ML and deep learning, which helped a lot. We can maybe write a dedicated blog post on our stack later!
open_source_enthusiast 6 months ago prev next
TensorFlow and Scikit-Learn, impressive! Any lessons on open-source dependencies in a startup environment?
startup_founder 6 months ago next
Sure @open_source_enthusiast! You want to always keep dependencies up-to-date, contributing to the projects if possible. It's crucial to have AI engineers on the team who know and respect the tools, otherwise you'll run into problems fast.
resources_minded 6 months ago prev next
I have been looking for free or low-cost NLP resources. Which resources did you use while building your solution?
language_creator 6 months ago next
Our favorite free resources for NLP are the GLoVe word embeddings (http://nlp.stanford.edu/projects/glove/) and Spacy package (https://spacy.io) for tokenizing and part of speech tagging. Hope it gives your resources a boost!
ai_explorer 6 months ago prev next
Fascinating insights on content filtering AI. What industry verticals have the most demand for your platform?
startup_founder 6 months ago next
@ai_explorer, so far we see demand mainly from education, media, and HR - but it can be applicable to a wide variety of industries. We're very curious to see who implements this technology in novel ways!
alternative_thinker 6 months ago prev next
Wondering if you also developed tools that preserve creator intent while filtering. What's your take?
corpus_master 6 months ago next
Yes! Understanding context behind content is essential to preserving intent. We built several techniques for this improvement in H1 2023: context-aware word embeddings, custom BERT models, and budget-friendly NLP libraries. Check out our latest paper: https://arxiv.org/abs/3002.11111