Harbor

Inspiration

Putting myself in the shoes of the victims, the first loud emotion I felt was fear: fear from the threats, and fear of the unknown. In that moment, more than any notification, I knew that having an expert at my side would reassure me immediately. That is what we set out to build.

What it does

Harbor is a personal security analyst for Sound Credit Union members. It protects against authorized scams and account takeover.

• Spots the moment of risk. Account events (password resets, contact changes, new recipients) quietly build a picture. When a payment starts, transparent rules decide: do nothing, ask a light question, or pause it.
• Talks like a calm expert. The member types anything in their own words. Harbor is honest, never blames, and keeps them in control. Takeover check-ins go to the phone on file before the changes.
• The AI never controls the money. The model reads the conversation and flags risk. Our code decides what options appear, so the model can't release, cancel or hold a payment.
• Hands off to a real person with the full context, so the member never repeats their story.
• Treats legitimate members fairly. Look-alike checks and a "Planned Big Payment" turn likely false holds into a light check or quick callback.

How we built it

• Trigger engine: Arm and Decide stages in JavaScript, tested on all 118 synthetic members in a live simulator.
• Prototype: a member phone app, an employee console, and a chart showing every rule check.
• Agent: a local Qwen 2.5 14B model (Ollama) behind a prompt, a reply filter, and code-controlled options.
• Backend: Kafka-compatible events, a rules scorer (about 1 ms), an advisory LLM tagger, a notification agent, Postgres, and a retro job for longer-window patterns.

Challenges we ran into

• Look-alikes. A simple takeover rule flagged all 4 legitimate "real move" members. They changed only their address and paid a day later.
• Amount alone can't tell scams from real payments. A $51K home-closing wire looks like a scam on the numbers.
• Making an LLM safe. Testing exposed shared passwords, pressure from relatives, and prompt injection. We fixed them in code, not just prompts.
• Honest evaluation. We tuned on the same synthetic data we tested on, so our results are in-sample.

Accomplishments that we're proud of

• Every anchor and look-alike story works end to end, including a scam we miss and a false positive.
• In-sample, the rules paused all 48 scam and takeover members and held 7 of 70 legitimate ones, all large wires or cash purchases. These are not real fraud rates.
• A working streaming pipeline, not just mockups.
• Safeguards for relatives or caregivers who may be the exploiter.

What we learned

• A quiet pause often beats a louder warning. In our data, warnings were clicked through every time.
• Honesty beats reassurance: members need to know they're in control.
• Use the LLM for understanding and writing, and keep money decisions in plain, testable code.

What's next for Harbor

• Run in shadow mode on real data to measure true false positives.
• Evaluate on held-out data.
• Add real SMS and push delivery, plus phone and branch channels.
• Test tone and accessibility with real members.

Built with

JavaScript, Node.js, Redpanda (Kafka API), PostgreSQL, Docker, Ollama, Qwen 2.5

Built With

• docker
• ollama
• postgresql
• qwen