Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile
Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท

@alex_picprod

Engineering leader. Prev @hertilityhealth ๐Ÿ‘ฉโ€โš•๏ธ and @thetrainline ๐Ÿš‚. Building simplesync.co and heycookbook.com

ID: 283476807

linkhttps://alexrieux.com calendar_today17-04-2011 11:00:57

1,1K Tweet

473 Takipรงi

440 Takip Edilen

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

Got some great result with using Langchain, GPT, function calls and JSON schema to extract recipes from web pages into a formatted JSON for heycookbook.com . No need to maintain any HTML parsing and formatting!

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

I have been using LLMs, Langchain and RAG for two use cases and I am really impressed about the results: 1) Extract and transform data into a structured format using with_structured_output 2) Perform text matching with embedding

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

My indie hacking breakdown lately: ๐ŸŽจ 40% crafting ads and landing page designs on Figma ๐Ÿ“ˆ 20% setting up analytics and ad campaigns ๐Ÿงช 20% researching lead acquisition strategies ๐Ÿ’ป 20% building with LangChain and OpenAI Check out my project at heycookbook.com

My indie hacking breakdown lately:

๐ŸŽจ 40% crafting ads and landing page designs on Figma

๐Ÿ“ˆ 20% setting up analytics and ad campaigns

๐Ÿงช 20% researching lead acquisition strategies

๐Ÿ’ป 20% building with <a href="/LangChainAI/">LangChain</a> and <a href="/OpenAI/">OpenAI</a>

Check out my project at heycookbook.com
Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

I am writing some Terraform to deploy a Python script to AWS lambda as a docker image, it turns out more challenging than I thought but cool to get 10GB function size limit, especially since those ML dependencies are really big!

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

Just discovered React Email to build emails using React and export to HTML that works with any email clients! This is very much welcomed to build out my marketing emails! Link: react.email/docs/introductโ€ฆ

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

After using GitHub Copilot for more than a year: - Code completion ๐Ÿ‘Ž - Copilot chat ๐Ÿ‘ Great to be able to prompt with current file as context but the code completion still sucks. Would be great to be able to choose the model, heard Anthropic โ€˜s model is great for code

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

Claude 3.5 from Anthropic is giving me much better results than GPT 4 on code prompts (Terraform, NodeJS Lambda and Python Lambda). You can try it now here: claude.ai/chat

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

I am trying to build a reliable RAG pipeline but I am struggling to pull relevant data from a my vector DB, anyone with LLM and RAG experience could help me?

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

I thought building a RAG pipeline was all about using an LLM to query chunks in a vector DB although I would never get the right context for my queries. This article explain an interesting approach by storing summarised data alongside the chunks towardsdatascience.com/leveraging-smaโ€ฆ

Alex R ๐Ÿ‡ฌ๐Ÿ‡ง๐Ÿ‡ซ๐Ÿ‡ท (@alex_picprod) 's Twitter Profile Photo

Why did Google release the transformers whitepaper to the public? It feels like they could have used this breakthrough to have a huge competitive advantage in AI.