EcoMetrics

Elevator Pitch
Our solution
System design

Inspiration

Our teammate, Anchit, worked in ESG investing for 5 years and understood that ESG scores are broken. Despite ESG investors spending millions of $ hiring consultants to do this research, ESG scores are not accurate and do not reflect public sentiment. For example, ESG rating agencies awarded EV maker Tesla with an ESG score of 37/100 against a score of 84 awarded to tobacco company Phillip Morris, which is completely non-sensical.

What it does

EcoMetrics scores companies on ESG, based on public perception. Investors focusing on ESG can utilize these scores to formulate their investment decisions, as these scores are real-time and more accurate. Also, unlike ESG consultants, EcoMetrics is much cheaper.

How we built it

We used twitter as a data source and then extracted ESG related posts in a time period. We then used sentiment analysis to assign scores on each post. We then aggregated these posts to get a company score. Amongst the tools, we used:

LangChain: For orchestrating data flows and integrating with Large Language Models (LLMs).
Flask / FastAPI: For backend web development and API creation.
Hume AI: For sentiment analysis and advanced emotional AI processing.
Streamlit: For creating an interactive user interface.
Replit: For development and deployment.
LLMs (Groq): For natural language understanding and generation.
Crew AI: For Multi-AI Agents Collaboration Please find our detailed tech stack here: link

Challenges we ran into

Twitter APIs to extract data are expensive. We found some free APIs, but they unfortunately took a long time to extract data. We had to find a way that would allow us to extract social media posts for our project, which was challenging. Ultimately, we used Langchain with SerperAPI to find social media posts on the internet. Unfortunately, this process took a long time, so we had to scale down our initial plans of also presenting a lot of visualization. One of the big challenges was also prompt engineering of AI agents, in order to get better results. Also, there were other challenges like token limitations with Groq.

Accomplishments that we're proud of

For our project, we had to integrate multiple moving parts, including social media posts, Hume, Groq, Crew AI and Streamlit. We worked together as a great team for integrating these parts, and are proud of that.

What we learned

We learned a lot on how to make best use of pre-built APIs/ infrastructure to create a large and complex project.

What's next for EcoMetrics

We have a working prototype ready. The next step is to go out to the market and sign up ESG investors to use Ecometrics. We also intend to raise a seed round to be able to integrate the paid Twitter APIs and develop more features.