PartnerUp

• 

  Partner Up Logo

Inspiration

We were really inspired by the evolution of the marketing business from the traditional word of mouth/storytelling to digital social-media management. As Influencer marketing is the new paradigm of marketing goods and services, we thought that this would be the most interesting space in.

What it does

Partner-Up aims to use Artificial Intelligence and open source Language Learning Models to develop a chatbot assistant that helps emerging brands to leverage the power of social media marketing.

When an emerging brand is interested in growing their online community, increasing online sales and marketing online, they may seek to partner with influencers or agencies in their business category. We decided to work in the most exciting domain we could imagine, social media. We designed this chat assistant to save brands the time invested in searching for the right partners, marketing agencies or influencers, to work with. Partner Up, as a chat assistant, would help these emerging brands find a preferred marketing partner, generate an introduction email letter, making the process of marketing much easier.

To use the voiceflow prototype, you will have to introduce the name of your brand, how much you think you will spend on marketing. After choosing your desired agency, the assistant will generate an email draft, with your brand name inserted into paragraphs as needed. You will also get contact details for the agency, helping your brand initiate the social media marketing process faster.

How we got here

Most of our early researched involved conducting a user survey with social media influencers. While we did not end up building a tool for influencer in particular, this survey was still a useful exercise in user research. our prototyping was done in different programs, including: notion, Figma, Uizard and voiceflow. We started a workspace in databricks, imported an open source dolly model to our notebook, but are still working on developing the skills needed to train this model for the Partner Up project.

We built our prototype chatbot with voiceflow using their openai gpt integration. It represents the idea and ambition of what we want to achieve with the eventual implementation of the open-source resources that we discovered throught this competition.

Challenges we ran into

We were able to set up our notebook on databricks and connect a google compute cluster to the notebook. We did some testing and troubleshtehooting within the notebook with some python scripts. We are able to converse with Dolly v2-3b within our databricks notebook. In order to get the best out of the model, it will need training and engineering with our own datasets. There is also the possibility of learning from our voiceflow prototype what prompts we should train our version of dolly v2-3b on. Secondly, we tried experimenting with other open source solutions like huggingface, but were limited on time needed to explore this environment.

Finally, we ended up reverting to our voiceflow prototype which is available for the judges to try out. We haven't given up on the use of open source tools like dolly from databricks, huggingface, or any other datasets that were graciously provided to the public. We will be updating our project and seeing how we can develop this chat assistant with open source tools available for us to use.

Accomplishments that we're proud of

This competition has been very valuable in helping us develop adequate knowledge on the landscape of AI+Data science amongst other topics that are involved in developing tangible artificial intelligence software. It was hard, but interesting all through.

As a designer, I can better communicate with datascientists and engineers in the field of machine learning/ data science and AI, more than I could at the start of this event. Our abilities to network with other participants, and collaborate on what idea to pitch has improved. The challenge of the task we set out to accomplish, building an AI chat assistant for emerging brands, inevitably helped us as a group in improving our skills in the area of research, creativity and problem solving.

These aspects are important for any project and we greatly value the opportunity to have participated amongst such talented developers, designers and creatives.

What we learned

We learnt about the various components that are required to build a working piece of software with open source tools and data.

With a clear vision for our chatbot assistant, we can build it if we keep learning and developing our skills.

What's next for Partner-UP

making updates to our criteria search UI, with a focus on making the user experience more value driven. Integrating open source tools with software development. We will take time to reflect on our current prototype progress and re-visit and update our databricks notebook. A much more custom version of the chatbot will be released to the public for review and testing.

Built With

• figma
• google-docs
• gpt
• notion
• openai
• voiceflow