Building ALXO at FirstCommit
By Anshika Pal, Team Arixen | #BharatBuilds #WeMakeDevs #AWSBharatBuildsTour #FirstCommitHackathon
A client says, "Can you add a quick login page?" A week later: "Could we do one more revision round?" Neither request feels big, but a few weeks in, a freelancer has done hours of unpaid work and has nothing to point to when it's time to invoice.
That's the problem we set out to solve at FirstCommit.
Our team, Arixen, led by Ilakkiyan J, built ALXO, an AI-assisted scope-creep detector for freelancers and small agencies.
What ALXO does?
You upload a client conversation (TXT, CSV or JSON) along with the original contract scope. ALXO then:
- Classifies every message as In-Scope, Scope Creep, Clarification or Off-Topic
- Calculates the cost of extra work in code, never by the AI, so prices are always exact
- Shows an audit ledger and chart of how far the project drifted
- Generates a change order, a client-ready email and printable PDF receipt
It runs on AWS: Amazon Bedrock for classification, DynamoDB and S3 for storage, and Amplify for hosting.
My role: QA and documentation
Our team was spread across the country, so clear documentation mattered more than usual. I focused on making sure ALXO was well-defined and well-tested.
- Specs and requirements: I wrote the product specifications and functional requirements, giving everyone one reference for how ALXO should behave.
- UI design system audit: I checked the interface for consistency across components, colors and themes, and shared my findings with the frontend.
- QA across 15 test scenarios: Real chats are messy, with stickers, emojis and sometimes illegible text. I tested all 15 scenarios to see whether borderline requests were classified sensibly and whether the change-order output stayed accurate.
- Playbooks and demo script: I prepared the execution playbooks and wrote the script for our demo video.
Key takeaways
- Requirements first. Even in a short sprint, a shared document saves time.
- Test the messy cases. AI features break in the gray areas, not on clean examples.
- AI judges, code calculates. Keeping the math out of the model made ALXO's output trustworthy.
- Learn from your team. Watching my teammates make architecture and AWS decisions under pressure taught me a lot.
Thank you
Thanks to the FirstCommit organizers and to my teammates Ilakkiyan and Manikandan for a great experience.
🔗 Live app:  https://main.dhbgp6utbowvg.amplifyapp.com/Â
🔗 Code:  https://github.com/ilakkiyan-j/AlxoÂ
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article