Software Engineer@MagicAPI
View MarkdownAfter my sudden departure from the previous company, I was going through a rough patch. One of my most trusted mentors recommended me to API.market, and I couldn't be more grateful. The referral led me to a team I genuinely admire. The work, culture, and people here have been incredibly motivating and fulfilling. I had a lot of ownership of the platform. Things like revenue, databases etc. were shared transparently.
My Work
I have been working at a lot of tasks, but to mention some of the major changes to which I contributed:
Code Clean up
After my onboarding my first task was to clean up the code with the best practice. The Co-founder mentioned this:
Shashank Aggarwal - CEO & Founder of API.market
Why is it so important to have good code and why are we focussing so much on code quality? It directly affects our revenue, operations and where we spend our time. I can not express the compounding effects of maintaining a good code base and a terrible experience of maintaining a terrible codebase that affects our mental health and indirectly our revenue.
We had over 240 errors and warnings, after the cleanup there were just 19.
Responsiveness for the platform over various devices
The mobile UI was not the most important thing we cared about as our customers were business people on desktops, but it had to be fixed. So I fixed and optimised the UI for mobile. Things like duplicate nav menus, I had to see and fix the components.
Re-Designing Seller Dashboard for better managing APIs
This re-design was majorly focussed to improve the usability of the seller Dashboard supported by API.market for the sellers to manage and visualize APIs and their revenues. In this work I improved the design of the dashboard using shadcn-ui and improved the responsiveness.
Added Wallet system for users to add and store funds
This was done by keeping some future things in mind like wallets will be used to store credits provided to users, also we were planning to add. So in the old way of transactions on API.market users had to add their cards and the card marked as default would be used to make a transaction. But after the introduction of this wallet system users can add money via cards or we add credits in their account(to be implemented in future) and use them in any transaction on API.market platform like this:
Adding Vouchers for users to redeem Credits in Wallet
During our product launch, we had a plan to give 100 vouchers worth $5 each to all the production people who came to our website. Additionally, when we had a hackathon on our campus, we wanted to provide vouchers for participants. With these requirements in mind, we developed a comprehensive voucher system where users could claim public vouchers and the money would be credited into their wallet, which they could then utilize on the platform.
We implemented two distinct types of vouchers to meet different use cases. The first type was a targeted voucher system where we could could claim the voucher by mentioning specific email addresses. This allowed for precise control over voucher distribution for specific events or user groups. The second type was a where people could freely claim it once per account, with a maximum limit of 100 people. This system provided provided flexibility for both controlled distribution and open public access while maintaining security and preventing abuse.
New Hero Section Design
I redesigned the hero section of the API.market website to enhance user engagement and provide clearer information hierarchy. The new design featured improved visual flow, better call-to-action placement, and more intuitive navigation elements.
Smart Onboarding form
I redesigned the smart onboarding form to enhance user engagement and provide clearer information hierarchy. The new design featured improved visual flow, better call-to-action placement, and more intuitive navigation elements.
Converting full time!!
At the end of June, Shashank asked if I wanted to continue full-time. I was confident in my ability to work at that capacity, so I immediately said yes. I had recently received a full-time offer and unexpectedly exited that role. The internship had begun around the same time but offered slightly less pay than the full-time position. However, it was a clear decision for me, as I received a decent raise.
Moving to Noveum.ai
Around the same time I got full-time, the company was moving to it's next venture on Noveum.ai. Noveum is a platform to evaluate AI agents and auto-fix them. It was very exciting for me to work on a company from the very ideation phase to build it from ground up. In the starting I was the only person working on the infra/tech side. I managed the complete kubernetes clusters and the architecture for building a system that was gonna handle millions of requests/day!
We didn't build everything in the next.js app from ground app, instead we had purchased a template, so it had everything like auth, multi-org login and onboarding flow setup. Some of the major parts of my work at noveum were:
Traces UI
I worked on the traces UI to make agent executions easy to inspect. Traces are basically an end-to-end timeline of an agent run (steps, inputs/outputs, errors and latency). So the traces ui included a main trace view, flow visualization, expanded flow graphs, and filters like date ranges so debugging production runs was actually practical.
Datasets UI
Datasets are curated collections of examples extracted from traces (inputs, outputs, and selected spans). They’re used to run specific scorers at evaluation time, so you can measure quality, compare changes, and catch regressions in agents/prompts/models.
ETL Job with Agent
The ETL (Extract, Transform, Load) pipeline converted traces into datasets via an asynchronous background workflow. I built the end-to-end job orchestration using BullMQ workers, deployed it to a Kubernetes cluster, and operated the Redis infrastructure it depended on.