JigsawStack Logo

Beta

This is why we are building JigsawStack

image

Problem

Part 1: Backend tasks

Backend developers spend ~80% of their time maintaining services like web scraping, OCR, translation, and more that fit into different pipelines of a backend.

A good example is web scraping, which breaks every other week due to layout changes, security updates, IP address rotation, or infrastructure cost problems, requiring devs to rewrite code frequently.

Now multiply this pain across the hundreds of services and tasks that make up a modern web backend.

Part 2: Current providers

Similar solutions by large cloud giants offer outdated models, horrible performance and are highly expensive as you scale. Developer experience (DX) is basically non-existent in many cases.

Solution

So we asked ourselves in the age of AI, why isn't there a SDK that developers can plug into their backend that solves their high-maintenance tasks automatically using models that can adapt with simple DX.

Solving it

Backend tasks requires high performance, consistency and reliability especially at scale with mission critical business running on them from health care to the military but with AI today we have high inconsistency due to hallucination, slow performance with GPU costing and stability issues which don't make great replacements for backend work.

So we built our own small models that are highly specialized for each of the specific tasks like OCR, translation, web scraping, etc. Each model recipe consist of 3 parts, a base traditional machine learning model, a small language model and tons of custom data we collected and tagged for training.

What we get is highly specialized models that are fast, consistent and is easily distributable at global scale for high reliability.

To end it off we packaged it all into a single DX friendly AI SDK that can be easily integrated into any backend.

Our mantra

We believe in a set of principles that guide us in building JigsawStack:

  • Developer first: Meaning always speaking to you with no ego before we make make any major decisions or release. From how we design our SDK to the inputs the models can handle. It's decided after speaking to hundreds of engineers.
  • Accessibility: No matter how big we grow, we want to make sure every developer can use JigsawStack.
  • Speed: We want to move fast like our models. Which means always we have to keep shipping and investing more into research implementation.
  • Love: We have to love what we do. Everyone single person on our team loves building which means pushing boundaries with quality even when its not easy.

Cheers, JigsawStack Team