Devpost

Inspiration

Imagine that you just received an offer to be a founding engineer at a seed-stage startup. Exciting, right?? 🎉

The thing is, now you need to set up a follow-up call with the founder to negotiate the compensation, and you have no idea how to have that conversation 😧

Not only do you have no idea what the market rate for a founding engineer is (because you've only worked in big tech), you also don't really know how to navigate the nuances of negotiation (what if you lose out on tens of thousands of dollars, just because you didn't know the right things to say in a 20-minute conversation??) 😨

This is the exact situation that I was in two months ago. And I had an especially big chip on my shoulder, because:

I'm a woman. It's a known fact that women are less likely to negotiate than men are. I was really concerned that I was going to get lowballed.
I'm Asian. Asian culture is a more communal culture, meaning that it places emphasis on maintaining the collective well-being over the well-being of any one individual. In other words, Asians are that much less likely to negotiate. I can confirm this from personal experience – when I got an offer from Dropbox, my dad straight up told me not to negotiate...at al 🤯 If I had listened to his advice, I would've been making $30k less on an annual basis.

That kind of stuff compounds. It's freakin' important to negotiate. And it's especially important for individuals from underrepresented minorities to negotiate.

What it does

Negotiation Nation is an on-demand chatbot that:

gathers existing data points (i.e. job description, salary range, your LinkedIn, etc.)
helps you roleplay the entire negotiation conversation
gives you immediate feedback on your mock negotiation conversations

How we built it

For the frontend, we used React, Next.js, and Tailwind CSS. For the backend, we used Python, Flask, Langchain, and the OpenAI API.

We used Unstructured.io to parse the data from the candidate profile and job description. This enables for a more realistic simulation of the chatbot experience.

Challenges we ran into

There were two directions we went back and forth between at the beginning: do we build a live teleprompter-like platform or do we build a chatbot that allows you to practice mock conversations?

There was a lot of debugging involved to iterate on the prompt, so that the responses from GPT-4 would be properly templatized to drive application functionality as needed.

Accomplishments that we're proud of

We ultimately decided on the latter, because we realized that it would make for a better overall user experience. Users probably want to walk into a salary negotiation conversation feeling super prepared, rather than walk into one with the intent of reading from a teleprompter.

Also, we're proud of our mission to drive equity in the workplace, starting with salary negotiation.

What we learned

It’s really important to build something with the user experience in mind.

What's next for Negotiation Nation

Eventually, we could totally build a teleprompter-like experience. However, we would assume that the user is pretty prepared, and we would only suggest things that would be the cherry on top (i.e. how to phrase certain asks).