From Zero to AWS: What I Learned Building Kabadiwala
How our team went from zero cloud experience to building and deploying a live e-waste platform on AWS, overcoming real-world user friction, TypeScript hurdles, and architecture bottlenecks.
From Zero to AWS: What I Learned Building Our Platform
By Aslam Sheikh, Team Payload (Pratik, Shruti, Aslam and Divyanshu)
Before this AWS hackathon, I had no real idea about AWS. I did not know what an EC2 instance actually felt like to deploy, IAM sounded like some random administrative acronym, and the console looked like a maze of endless options. I was stuck at the start and did not know where to begin.
A few weeks later, an app I helped build and stabilize is running live on it. This is my story of how that happened.
WHAT WE BUILT
Our team built a web application that connects local waste collectors with certified electronic recyclers.
We picked this problem because most of us give old chargers, broken cables, and drained batteries away without thinking twice. Electronic items contain valuable metals worth recovering, alongside hazardous substances that pollute when improperly discarded. Local collectors are simply earning an honest living, but they have no direct bridge to authorized recyclers and no way to verify fair market rates. At the same time, formal recyclers struggle to source volume from unorganized collectors. Both sides exist in silos, but nothing connected them.
In our app, a collector lists his collection items, uploads a photo, and enters the weight. He sees certified recyclers nearby with transparent per-kilo rates and picks one. He receives a digital handover code and QR code. When the recycler meets him, that code verifies the transaction in real time. Once payment settles, an automated receipt is emailed to him. The platform supports English, Hindi, and Marathi, and every recycler is verified so collectors only deal with genuine buyers.
WHAT THE SHOP OWNERS TAUGHT US
We did not want to build just sitting behind our laptops, so we stepped outside and spoke to local collection shop owners in the market. It was an eye-opener. One owner would not even answer our questions at first, but after spending time explaining our intent, he opened up—and we are genuinely grateful for that.
What he shared grounded our initial thinking:
- They want zero friction with tax paperwork.
- They already have established bulk buyers in other cities.
- They see little reason to switch unless an app saves them time and increases their earnings.
It was humbling, but it taught us a crucial engineering lesson: you cannot build software in a vacuum and expect people to adopt it. The UI had to be simple, ask for the bare minimum input, and never get in the way of a transaction.
MY PART
When you have a fast-moving hackathon team building different components simultaneously, things inevitably run into issues. My primary focus was frontend reliability, routing architecture, and UI/UX polish—taking the raw prototypes and making them feel like a cohesive, production-grade web application.
Here is what I worked on:
- UX and User Flow Review: Before touching components, I performed a wireflow review to trace how users navigate the platform. I mapped out friction points, confusing steps, and broken transitions to ensure both journeys felt natural.
- Routing Architecture: The dashboard routing was initially fragile, suffering from duplicate rendering issues where recycler views conflicted with collector routes. I refactored the router using React Router into a modular, reusable layout, cleanly separating role-based dashboard views.
- Navigation and Component Redesign: I added a persistent sidebar, expanded the header navigation, and redesigned the contact and profile sections. I also ensured the default authentication landing page cleanly pointed to sign-in for returning users.
- Authentication Theming and Dark Mode: Our Clerk sign-in and sign-up flows had contrast issues and theme glitches. I built a custom appearance configuration that integrated Clerk cleanly with our application theme.
- Guides and Repo Setup: I wrote clear setup instructions and testing workflows so anyone on the team could clone, build, and run the project without running into missing dependencies or configuration errors.
LEARNING AWS FROM ZERO
Coming into this, navigating cloud infrastructure felt like reading a foreign dialect. My routine was straightforward: break down complex architecture concepts with AI into plain human terms, test the configurations hands-on in our staging environment, and iterate until it clicked.
Seeing our cloud stack come together was one of the most rewarding parts:
- EC2 (Mumbai): Hosts our full stack in the ap-south-1 region for low latency, as our user base is based in India.
- IAM Roles: Configured so our backend server accesses AWS resources securely without storing secret keys in the repository.
- Amazon SNS: Triggers instant alerts when users submit support or query tickets.
- Amazon SES: Dispatches digital transaction receipts to collectors right after handover verification—giving small collectors verified proof without forcing them into paper ledgers.
- Amazon CloudWatch: Collects live server metrics and application logs, saving us hours of guessing when backend routes threw runtime errors.
- AWS Backup and Budgets: Scheduled automatic daily snapshots, while automated budget alerts kept us protected against accidental student credit overruns.
Once you realize every service solves a specific operational bottleneck, AWS stops being intimidating and starts feeling like an empowering toolbox.
THE BUGS
Building this was not smooth sailing; we tackled bugs at almost every layer:
- TypeScript and Build Bottlenecks: At one point, our frontend refused to compile due to numerous typing errors and lint failures scattered across inputs, category selections, weight estimators, and profile forms. I led a deep cleanup overhaul across more than 20 files to enforce strict type safety and get our build pipeline green.
- Item Creation Glitch and Weight Calculations: A runtime error broke the client whenever a collector attempted to save an item entry. I traced the state mismatch in the weight-estimation handler, patched the payload formatting, and stabilized the submission loop.
- Duplicate Dashboard Render Loops: Our collector and recycler dashboard routes kept colliding, causing multiple parent containers to mount on top of each other. Restructuring the nested route outlets resolved the layout jumping.
- Camera and Model Constraints: The camera scanner refused to initialize on mobile devices over insecure HTTP, requiring us to set up a reverse proxy with automatic SSL certificates. Furthermore, when foundation model access was restricted on our fresh AWS tier, we adapted the image verification pipeline with modular fallbacks.
Finding and eliminating these edge cases taught us how to debug methodically under pressure instead of patching blindly.
THE BIGGEST LESSON
The biggest takeaway from this journey was not just learning cloud architectures or code typing patterns. It was learning team execution and trust.
Knowing how to communicate early when a build fails, handing off tasks cleanly between backend, design, and deployment, and keeping each other motivated through late-night debugging sessions—that collaboration is what turned our project from an idea on paper into a production deployment.
Pratik, Shruti, and Divyanshu—thank you. Working with you made every single bug worth tackling.
WHAT'S NEXT
Our next milestone is a micro-aggregation hub model where neighborhood collectors can deposit smaller daily loads at local collection centers, pooling them into high-volume freight lots for formal recyclers. This significantly cuts logistical overhead per kilogram. Before writing the code, we plan to take our live prototype back to the market to validate the workflow directly with real users.
FINAL THOUGHTS
I walked into this hackathon with zero practical AWS experience. Today, we have a deployed, resilient platform running live on the cloud, a stable codebase, and an incredible team. If you have been putting off entering a hackathon because you feel you have not mastered the tech stack yet, just take the plunge. You will learn more in two weeks of building and fixing real systems than in months of tutorials alone.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article