Connect

• 

  main user page

What is RizzNet?

Talking to people is a vital skill, one that we use everyday and plays a big part in our success in life. Making friends, job networking, dating and more can all be made smoother with solid communication skills. Most people inhibit these skills, but let them lay dormant inside barbecue they lack the confidence to use them. We've all been there when you think of the perfect joke, pick-up line, icebreaker or cold-approach but didn't end up using it because we lacked the confidence to go for it.

Confidence can be built in a multitude of ways, however, one tried and true method is practice. Whatever the skill you are trying to build, enough practice can make you a master at it. Even better if you can get some feedback, and make your environment as similar as possible to the real thing. Then, over time whether you notice it or not, people around you will. You can be the smooth talkers from the movies, or the gentle-person whom chooses their words wisely.

That's why RizzNet is here! Put the silly name aside and have a look at what is does: Every day, you have access to one conversation with a unique customized AI agent, catered to what you are practicing for, where your aim is to entertain a conversation with whatever goal you choose. Want to get their number? go for it! A job? give it a go. However, the AI will be tracking how comfortable it feels throughout it's talk with you and can leave at any time if it feels it needs to. Once the conversation is over, feedback from throughout the conversation and an overview will be provided, alongside general suggestions and rating so that you can improve over time. The messaging system is entirely done through voice control, with a visual stimuli provided initially for context and salient para-linguistic features being displayed too, to provide users with as reasonable of a realistic environment as possible so that the practice is practical. This is reinforced by the singular attempt allowed per day, which means you can't back out of a conversation unnaturally as you do not get another go, putting needed real-life-simulating pressure on the user.

How we built it

RizzNet is a web app built with nextjs, with a neon database connected via drizzle ORM to store chat scenario data generated and used by the python flask API backend which uses GPT prompts to provide the chat with it's scenarios.

Challenges we ran into

Our team was unfortunately half the size that we initially planned, meaning the initial scope was likely unreasonable. Given that problem we adjusted ...and still ran into a lot of debugging with the tech stack, styling properly, refining the prompts to get reasonable outputs and more. Our schedules also didn't align very well for this project as we only had about one day each to work on it pushing us to some long hours crammed at once.

Accomplishments that we're proud of

Although the final product does not function how we imagined, there are a lot of nearly finished parts that if polished and combined together would resemble close to the initial project we imagined. So, despite the lackluster final product, we ..behind the scenes know the work that was put into it, the amount of docs read, lines wrote and hours spent that we committed to this.

What we learned

Firslty, we need to organize our team better, making sure we have enough members, and a better variety so that all aspects of the program can be handled smoothly, simultaneously. Additionally, before any future projects, some practice definitely needs to be done to spend less time reading through documentation. This project really showed us how much better we could've been with some practice (ironically ik lmao) Finally, it worth noting, although we tried to keep the scope smaller, in the future we should just try to produce the mvp, nothing else, not worrying about expand ability, or efficiency as much as we did. Our project would've benefited a lot from this type of ground up sprint-style approach.

What's next for RizzNet

Like mentioned previously, a bit of re configuring the code base with polished files to bring everything together and a solid prototype can be finally constructed. Going forward we would like to:

• Implement as many immersion features as possible (without taking away from the experience), including better and more image generation, and improved context providing
• refine the models to replicate real life experiences more accurately via community and professional help, such as interviews with target demographics to match experiences.
• refine models to encourage better behavior (discouraging mano-sphere and similar habits which may negatively impact society unknowingly to the user)
• add more models to choose from and improve user experience
• attempt some beta testing with a range of demographics to refine program

Built With

• api
• clerk
• drizzle
• gpt
• nextjs
• python
• tailwind
• typescript