Hi, I'm Akansha.
I'm a software engineer. I spent four years on the frontend, and I'm now going deeper into backend and distributed systems. What ties it together is making complicated systems understandable, whether that's a tool for engineers working through thousands of data sources or a cache whose metrics show what it's doing.
I studied Computer Science at IIIT Gwalior. Before graduating I interned at Rakuten, building an internal testing tool for QA engineers. At Reliance Jio I then led the move of JioMart's B2B partner app from native Android to the responsive web, and set up its design system.
Alongside Jio, through the LFX Mentorship, I added groupcache as a cache backend to Thanos, a CNCF project, in Go. It taught me to find my way around a large codebase fast, and to learn from how experienced maintainers review code.
Then came four years at Phaidra, an NVIDIA-backed startup building AI agents for data-centre operations, where I was the first frontend hire. I rewrote the portal's main chart, owned an internal data-mapping tool, built versioned editing on the control side, and chose the charting library for an AI feature whose charts hold a million points per series. Along the way I wrote PRDs and test plans, reviewed API specs and took on-call shifts.
Since Phaidra I've been going deeper on distributed systems: reading Designing Data-Intensive Applications and working through Fly.io's Gossip Glomers challenges in Go. I live in India, and I'm open to full-stack, backend and distributed systems roles, remote or hybrid.
How I work
Four habits I bring to a team.
Start from the ambiguity
My first project at Phaidra came with no spec and no agreed idea of how it should work. Much of the job was going back and forth with the people who’d use it until we agreed, and only then building. Later I wrote PRDs myself.
Make the state visible
A problem should be something you see, not something you hunt for. I like interfaces that show what a system is doing, not just whether it passed.
Design for the real data size
When a picker crashed on 10,000+ items, virtualization was the first fix and paginated server-side search the lasting one. I chose AG Charts over Highcharts for the same reason: charts had to stay smooth at a million points.
Use AI tools, keep the bar
I use Claude and Cursor every day to move faster on implementation, tests and docs. They don’t replace understanding what I ship.
How this site is built
Next.js and React, in strict TypeScript, styled with Tailwind CSS. Headings are set in Bricolage Grotesque, text in Figtree, and labels and dates in Geist Mono. The colours are matcha and sakura, in light or dark. The source is on GitHub.