Project Info

BlackMarkIt

Inspiration

In the Gulf of California, the Totoaba’s swim bladder is known as "aquatic cocaine," fetching between $20,000 to $80,000 per kilogram. This astronomical value has drawn in the many Cartels, transforming local fishing villages like San Felipe into nodes of a global criminal network, and a similar story is being told for many other endangered species today. So we built BlackMarkIt, a predictive modeling program that allows police departments to find traffickers before they reach a consumer.

What it does

BlackMarkIt is a global illicit-trade intelligence engine. It ingests seizure data (Origin, Transit, Destination, Commodity Type, etc) to map the hidden "connective tissue" of this $23B illegal wildlife market. Users click on regions our algorithm have determined to be likely spots for a black market to operate which starts a search algorithm for potential hubs (buildings, warehouses, stores, etc) and uses an agentic AI LLM agent to comb through online resources to check what companies are associated with what properties, whether these companies are shells, and their online presence. Through this, we determine which sites might be very likely within an area to be an operating base for these trafficking rings.

How we built it

We utilized a global dataset of police seizures (intercepted packages containing ivory, rhino horns, etc.). This data provided not just a "point on a map," but the economic lineage: monetary value, origin point, and interception point. We then fed this information into a DBSCAN algorithm that allowed us to create clusters with centroids that could act as potential market areas. Then by drawing a sweep around all the points in a given cluster, we could determine an area we deemed likely to house an operating base for a trafficking ring. Then with this area, we would iterate through all of the buildings by utilizing OpenStreetMap. For each buildings, we would then use a ChatGPT based agent we created to operate on an OSINT based framework to find information about the buildings and their handlers with publicly available resources, This agent also determines the risk of a given building, determining it to be more or less likely of being apart of a ring.

Challenges we ran into

One of the biggest challenges we faced was determining how we would cluster our points together. Initially, we were planning on creating a polygon based model where all the interception points act as a vertex of a given polygon area that we would deem as an area with higher probability of having a black market. This was later adapted to use a multidimensional k-means algorithm that weighs position, weight, and cost against one another to determine clusters from which their centroids would be the hotspots. We later had to pivot this idea once again as kmeans would have issues in how groupings were determined as well as how we created the area containing a possible black market. In the end, we decided to utilize a DBSCAN which would comb through the data and create groupings of closely knit points and have these clusters become the new searchable areas and had a lot higher accuracy.

Accomplishments we're proud of

We were really proud of the fact that we were able to create a fun and unique idea product during the limited time of this hackathon. We wanted to incorporate a lot of technical and aesthetic features that would still streamline user interaction with the platform.

What we learned

We learned about utilizing unique datasets with multiple non quantifiable metrics while still being able to train a prediction model upon it. We also learned how to leverage agentic AI to aid in searching that would have taken a person much longer to complete.

What's next

Real time endangered species commodity package determination Transition into using models to find hubs for human trafficking and drug distribution or other hidden markets.

Development Setup

Backend

Create and activate a Python virtual environment (recommended):
```
python -m venv .venv && source .venv/bin/activate
```

Install dependencies:

pip install -r backend/requirements.txt

Start the FastAPI server (from the backend directory):
```
cd backend
fastapi dev main.py --reload --port 8000
```
The API will be available at http://127.0.0.1:8000.

Frontend

Install Node dependencies:
```
cd frontend
npm install
```
Run the development server:
```
npm run dev -- --host 0.0.0.0 --port 5173
```
The React app will be served on http://127.0.0.1:5173.

Notes

If both servers are running, ensure the frontend points to the backend API (check src for proxy or fetch endpoints).
Use npm run build (frontend) and uvicorn --reload (backend) for production simulation.

Analysis

Compare with all teams

View

Metric

sathvikr
19
pvelleleth
12

Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.

Technology

Found in codeClaimed only

CSSIn code
FastAPIIn code
HTMLIn code
JavaScriptIn code
OpenAIIn code
PythonIn code
ReactIn code
Tailwind CSSIn code
TypeScriptIn code

9 of 9 appear in the indexed code.

AI coding agents

CodexConfig

Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.

Codebase size

Source size

545 KB

Source files

108

Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.