ML Under the Hood

ML Under the Hood is a newsletter that examines the development and implementation of machine learning (ML) and Large Language Models (LLMs) in the context of creating ML-driven software products. It revolves around technical insights, project management strategies, and industry trends, focusing on practical applications like ChatGPT, open-source alternatives, and effective software engineering practices. The newsletter also discusses the importance of focus on user experience, problem-driven design, and the balance between technological perfection and shipping functional products.

Machine Learning Models Software Engineering Practices Project Management Open Source Contributions Productivity and Personal Development Technical Implementation Details Privacy and Analytics Cost-effectiveness in ML Deployment Industry Trends and Collaborations User Experience Focused Development

The hottest Substack posts of ML Under the Hood

And their main takeaways

On keeping MVP ugly and personal productivity systems

0 implied HN points • 07 Feb 22
  1. Keep your MVP (Minimum Viable Product) simple and functional.
  2. Implement a solid backup process for important personal data.
  3. Develop a personal productivity system that includes journaling and flexibility.

What are ML Pipelines and how do they come to be?

0 implied HN points • 12 Feb 23
  1. Machine Learning focuses on automating specific tasks through ML models.
  2. ML Pipelines are workflows that transform data into trained ML models.
  3. ML projects start with prototypes and evolve into standardized workflows for team collaboration.

Black Friday code is open source, ChatGPT assistant and its future

0 implied HN points • 05 Feb 23
  1. Black Friday code implementation is open source on GitHub for community review and contribution.
  2. ChatGPT is a helpful assistant for generating code snippets, APIs, and UI components instantly.
  3. The future of Large Language Models looks promising, with open source projects potentially making them more accessible and cost-effective.

Comparing Event-Sourcing in F# and go - getting unstuck to ship that code

0 implied HN points • 28 Jan 23
  1. The importance of shipping code rather than getting stuck in the pursuit of perfect performance
  2. Golang is useful for complex business domains and SQLite is praised for its speed and fault-resistance
  3. Tech tools are helpful, but ultimately what matters most is shipping the project
Get a weekly roundup of the best Substack posts, by hacker news affinity:

Deterministic simulation, Obsidian and opinionated tech - Abdullin.com

0 implied HN points • 24 Sep 22
  1. Blog migration to Obsidian backend is complete with new content
  2. New articles on deterministic simulation in software and event sourcing series available
  3. Interesting discoveries include a new way of blogging about Golang, simpler than react HTMLX, and Python's built-in persistent key-value store

Abdullin.com: designing lean analytics pipeline for a website

0 implied HN points • 19 Jun 22
  1. The website removed Google Analytics and now uses access logs without cookies or IP tracking for better user privacy.
  2. The new data architecture focuses on lean analytics and privacy-first design.
  3. The changes are already implemented, showcasing a production-ready lean analytics pipeline.

Fine-tuning for GPT-3.5

0 implied HN points • 23 Aug 23
🕹 Technology AI Cost Specialization Documentation
  1. This fine-tuning makes the model specialized for specific tasks, not to learn new things easily.
  2. Training costs for the fine-tuned GPT-3.5 are relatively low at $0.008 per 1K tokens.
  3. Using the fine-tuned model is 8x more expensive than the base GPT-3.5-Turbo.

Projects where I focused on wrong and right problems

0 implied HN points • 10 Jan 22
  1. One project went wrong as I focused too much on perfecting the tech stack.
  2. Another project succeeded by ignoring tech and prioritizing user experience.
  3. It's crucial to focus on solving the right problems to achieve successful outcomes.

Problem-Driven Design and DDD

0 implied HN points • 02 Jan 22
  1. Software engineering involves designing prototypes and building solutions.
  2. Problem-Driven Design emphasizes the importance of focusing on the problem at hand.
  3. DDD and Problem-Driven Design have differences that are worth exploring.

Underdog joins the fight

0 implied HN points • 05 Oct 23
🕹 Technology AI Cloud Models ML
  1. Anthropic partners with Amazon in a $4B deal, offering access to second best LLM model through an API on AWS Bedrock
  2. Cloudflare introduces Workers AI to run low-power LLM models worldwide, aiming for data localization compliance
  3. Mistral AI releases a powerful 7B model with Apache 2.0 license, outperforming larger models and providing true open-source capability

Friends, Romans, countrymen, lend me your ears

0 implied HN points • 25 Feb 23
🕹 Technology Development Models Backend Hardware
  1. Developing a prototype ML product for niche languages and cultures has unique challenges that are not present in more common languages.
  2. Focusing on core objectives is crucial for efficient development and achieving sprint goals.
  3. Prioritizing functionality over speed in ML inference pipelines can lead to tangible progress and real product advancements.

Breaking the curse of LLM v2

0 implied HN points • 18 Jul 23
🕹 Technology AI Models Updates Performance
  1. New releases of large language models focus on efficiency over quality
  2. Performance improvements in GPT-4 and other models may sacrifice quality in some tasks
  3. LLaMA v2 by Meta offers better quality and commercial use but comes with language limitations and user restrictions

Writing things down and beautiful technical debt

0 implied HN points • 23 Jan 22
  1. Writing things down is valuable when working on multiple projects simultaneously.
  2. Technical debt can be seen as beautiful and can help in navigating complex landscapes to build better products.
  3. Documenting information and acknowledging existing technical debt are important aspects in project development.