Here’s a short demo of the chatbot in action. This walkthrough shows how the prototype responds to juror questions using public legal references. It's not flashy, but it works.
The name was a joke at first.
“Your Honor, I Object (to Jury Duty).”
But somewhere between running the repo and rewriting legal prompts, I realized I was doing more than tinkering.
I was building a real prototype.
The chatbot began as a weekend experiment. A stretch project that sounded just out of reach, but close enough to try. I had been exploring generative AI tools in the abstract, working through AWS SkillBuilder courses and watching repo demos. This time I wanted to go beyond theory. I wanted to build something tangible. Something real. Something useful.
I started with Ryan Gertz’s open-source repo. It was a clean, simple chatbot powered by AWS Bedrock and Streamlit. His instructions were solid, the codebase was public, and it used Claude as the backend model. It also looked like a puzzle missing its instructions.
Getting it up and running in the AWS Console felt like cracking a safe without knowing how locks work. It took trial and error, and more than a little help. That’s when I turned to ChatGPT. Not just for answers, but for partnership. At times I needed a collaborator. At other times I needed a guide. Sometimes I just needed a transparent tool. ChatGPT became all of those things. It was my sixth person off the bench, showing up with documentation help, error debugging, and quiet reminders like, “Did you activate your virtual environment?”
From there, I adapted the app into something new. A jury eligibility chatbot using real, public legal content from California Code of Civil Procedure sections 190 through 242 and Rule 2.1008 of the California Rules of Court. The prototype is cautious by design. It gives neutral, structured responses to common juror questions. It includes references to legal text, avoids interpretation, and tells users to call the jury office for edge cases. It is hosted locally, not deployed. No proprietary court data was used. This was a personal learning project, nothing more, nothing less.
Still, building it was harder than I expected. Mentally, I was taxed. I had to juggle Python, the AWS console, embedding models, and an unfamiliar development environment in VS Code. The vocabulary alone was a climb. FAISS, vector index, semantic similarity, inference parameters. I still don’t fully understand why I need folders like _pycache_/ and .venv/, but I know the app doesn’t run without them. That was part of the lesson.
I also struggled with perfectionism. I kept wanting to improve the bot. Tweak the responses, make the phrasing more natural, upgrade the user experience. At some point, I had to remind myself what this was. A proof of concept. Not a product. This wasn’t about polish. It was about process.
And the process worked. Here’s what I built:
🔗 Your Honor, I Object (to Jury Duty) v9 – GitHub Project
Here’s what I would like to add in version 10:
Plain-language summaries for each legal response
Multi-turn memory for natural back-and-forth conversation
Semantic search for better matching across phrasing
Exception handling for unclear or vague queries
Live deployment to secure internal or public-facing sites
Integration with court FAQs or local rules
Jury Office contact info in fallback messages
Prompts for unique situations like “Call the Jury Office”
Clear, actionable next steps for deferral, excuse, or disqualification
Safety guardrails and improved refusal handling
Tuned inference parameters for a neutral, cautious tone
But the most important thing I built wasn’t a chatbot. It was momentum.
The turning point came after reading a piece by Dr. Gemma Leigh Roberts titled “What If You Didn’t Wait to Feel Ready?”. She writes, “Confidence often follows action.” And later, “We build readiness and confidence by doing.” That stuck with me. Because that’s exactly what happened.
I didn’t wait to feel ready. I built something small, meaningful, and complete. In doing so, I gave myself proof. Not of mastery. Not of expertise. But of capacity.
That proof is why I’m going to CCC AI Summer Camp. This project gave me a head start. It also opened my eyes to what’s possible when you stop watching and start building. A few months ago, I was studying data analytics. Now I am coding a chatbot. Not because I had the credentials. Because I had a reason. And because I had help.
So no, this isn’t a launch-ready product. It is a prototype built in one weekend. But it is also proof. Not of perfection, but of possibility.
AI-assisted, but human-approved, just like any good front office move. Chat GPT the sixth person off the bench editor for this post. Every take is mine.