CounterMarket

Inspiration

Women still earn less than men, and the gap is even larger for women of color. The hardest part is not just the numbers, it is that most people negotiate without a reliable baseline. CounterMarket was inspired by the idea that pay transparency should be usable in the exact moment someone is deciding whether to accept an offer or push back with confidence.

What it does

CounterMarket helps users understand their market value and negotiate fair pay using data and AI. It compares a user’s salary to market percentiles, visualizes pay gaps by gender and ethnicity, and adjusts expectations using cost of living by metro area. It also generates negotiation scripts and provides an AI salary advisor so users can turn analysis into action immediately.

How we built it

We built a React frontend with Tailwind and Recharts for charts, and a Python Flask backend that serves analytics endpoints. Salary records and the negotiation knowledge base live in Snowflake, where we compute market percentiles using Snowflake SQL. For AI, we use Snowflake Cortex with Llama 3.1 8b for responses, and E5 base v2 embeddings for semantic retrieval. Our RAG flow embeds the user question, retrieves the top relevant negotiation articles using vector cosine similarity, and then generates grounded advice and scripts using the retrieved context plus the user’s market results.

Challenges we ran into

We had to move fast on environment setup since .env files cannot be pushed to GitHub, and every teammate needed the correct Snowflake and Supabase configuration. We also worked through frontend to backend issues like proxying, CORS, and authentication edge cases that can cause a blank screen if keys are missing. On the AI side, the biggest challenge was getting RAG to feel genuinely helpful, which required tuning retrieval and tightening prompts so the advice stayed specific to the user’s percentile and target range.

Accomplishments that we're proud of

We shipped a working end to end product that goes from salary input to analytics charts to an AI powered negotiation output. We are especially proud of using Snowflake not only for storage but also for real analytics and Cortex powered RAG, so the AI outputs are grounded and connected to the data. We also focused on a clean UI that makes pay gap insights easy to understand within seconds, which matters a lot during judging.

What we learned

We learned that percentiles are the simplest way to explain compensation because they create instant context, even for users who do not know market numbers. We also learned that RAG makes AI feel far more trustworthy because it can pull from a controlled knowledge base instead of improvising. Finally, we learned that reliability matters as much as features in a hackathon, so handling missing keys, stable local ports, and clear setup steps can make or break the demo.

What's next for CounterMarket

Next, we want to expand the dataset and add confidence scoring based on sample size so users understand how strong a comparison is. We also want exports like email and PDF negotiation packages, plus a hosted deployment so anyone can try it without setup. Longer term, we see CounterMarket as a financial wellness tool that can integrate into career platforms or banking apps to help users raise income at the moment it matters most.