Results of the first Daisi Hackathon!

Daisi (app.daisi.io) is the go-to community platform for Python Cloud functions and apps, deployed in a serverless fashion. At Daisi, we are on a mission to advance the deployment of incredibly impactful and game-changing algorithms or ML models. We want them to be readily actionable so everyone can start using any of them with simply one line of code.

Our first hackathon aimed at seeding the Daisi catalog with ready-to-be-called cloud functions in the following categories:

Back to basics, Daisies can be at the re used in many worflows, for instance

• Data access: interface with a valuable and often accessed database
• Data processing or transformation (Image formatting, documents formatting, paragraph extractions, data conversion from a format to an other…)
• ML models which can be used as “Lego bricks” to build applications, e.g., segmentation, classification, question answering

Game changers

Daisies can unlock new domain-specific capabilities and workflows, e.g., Geoscience, climate change, or astrophysics. Examples of these include:

• Domain specific algorithms or models, either novel or complex enough and not available in common packages
• Unlock access to useful data or interesting datasets that might not be directly accessible (including hard-to-find datasets)

Building the future

Daisies can deploy novel, curious, and innovative code! For instance:

• Internet-of-things
• Web 3.0
• Long term science for climate change, energy consumption, etc…

Submissions have been evaluated in term of usefulness (3 best Daisies,Jury’s choice), usage (first two Daisies with the most API calls) and popularity (first two Daisies with the most stars).

More than 100 Daisies have been submitted during the hackathon. Selecting only a handful of them among the high quality submissions has been a tough challenge. The choice made here reflects the diversity of the submissions and is an illustration of the creative use of the platform made by the participants.

Here are the winners:

Judge’s Choice

1st place : Shipment Delivery Optimization (by soul0101)

Daisies are serveless microservices which can be deployed instantly based on any Python function. Moreover, combining multiple Daisies to create a microservices backend is incredibly straighforward.

The “Shipment Delivery Optimization” Daisi leverages this capability and demonstrates how to build a sophisticated application dependant on two other Daisies which deploys a service for itineraries calculation and optimization. It even offers a large range of options for search and optimization algorithms. It is ready to be called and leveraged by any webapp needing a routing optimization service!

2nd place : Image Super Resolution (by oghli)

SISR (Single Image Super Resolution) refers to the task of enhancing the resolution of an image from low-resolution to high. It has applications in a number of domains including surveillance and security, medical imagery, reducing storage costs or increasing the performance of OCR systems for instance.

The “Image Super Resolution” Daisi is the deployment of a model originally published by Liu et al. in “An Attention-Based Approach for Single Image Super Resolution” and hosted on OpenVino. This model is now available for inference with only one line of code using the pydaisi package!

3rd place : Crypotcurrencies financial metrics (by jotatd)

As crypto currencies are getting more widely adopted, a fast and straightforward assessment tool for comparing the return and volatility of two crypto currencies is essential.

The “Crypotcurrencies financial metrics” calculator makes it incredibly easy. It implements a series of calculations on the top of the price history of two currencies and make them only one command away!

Most API calls

1st place : Multilingual-Indian Language Abuse detection (by kanav)

India is a multilingual society with people having a number of languages spoken across the country. With a total of 1.3 billion people and 10 languages used across different states identifying/reporting abusive content on social medial platforms is a growing challenge.

The Multilingual-Indian Language-Abuse-detection Daisi is the deployment of a language model trained on millions of rows having abusive content across different Indian languages and which has been called over 2.4M times during the hackathon timeframe!

2nd place : Interface to the Ethereum blockchain (by yogeshbansal)

This Daisi simplifies how to retrived data from the Ethereum blockchain. It provides a pre-authorized access using Infura Node and without hosting any of the infrastructure related to ethereum node, which is not possible on consumer grade hardware.

As it definitely breaks a barrier for users to interact with live blockchain data, it has been called over 1.5M times during the hackathon!

Most stars

The two winners in this category are the public choice, since what is rewarded is the number of “stars” given by registered and verified users in the Daisi platform.

1st place : NLTK for Tamil (by vijay)

Tamil is one of the oldest language in the world, spoken today by more than 80 million people. Still it lacks support from the modern Natural Language Processing toolkit.

The NLTK for Tamil Daisi contributes to the democratization ofNLP by making common tasks such as stemming, sentiment analysis or summarization available for Tamil vocabulary and sentences available globally, using the app or the API that the Daisi provides! It is widely popular and has been the most liked.

2nd place : Stock Market Analysis Using Machine Learning (by farhun)

The Stock Market Analysis Using Machine Learning Daisi democratizes techniques for stocks analysis by deploying a comprehensive app on the Daisi platform to understand and compare the historical performance of various stocks. Also a very popular Daisi, which comes second in term of number of likes.

Congratulations to the winners and we want also to recognize all participants who have created incredibly useful and valuable Daisies to be be discovered and re used by everyone. Your work is an inspiration for the community and you contribute to make the Daisi platform (app.daisi.io) a go to resource for the deployment and discovery of actionable Python cloud functions.

Building on the success of this first hackathon, we are continuing the hackathon with a Round2, deadline on Oct 16th. We will welcome submissions in a number of technical categories. And the Daisi platform offers now serverless GPUs for the deployment and fast inference of ML models! Look at the detail on the hackathon page and submit impactful and transformative Daisies in one of the following categories:

1. Document AI
2. New trends in Machine Learning. Interface with GPT-3, Dall-E, or similar
3. ETL Daisies
4. Remote sensing Daisies
5. Science ML Daisies
6. Robotics and IoT
7. Crypto, NFTs, and Finance