logr.it/koushith
Product engineer with a design background, working across the developer-tools and product stack. Currently leading core tech at Consuma AI; previously on the founding teams of Reclaim Protocol and Consenso Labs. I care as much about how software feels as how it works.
I posted a note that stuck with me while building AI products: I'd underestimated how much engineering goes into making an AI system feel simple. The cleaner the UX, the messier the backend behind it.
I joked about my new 'coworker' — a little desktop-pet cat that doesn't review PRs, join standups or ask for Linear updates, just follows my cursor around the screen and occasionally says 'meow'. A 10/10 teammate.
I wrote up why Vortex died: rendering live third-party sites inside a canvas meant fighting Manifest V3, CSP frame-ancestors, X-Frame-Options and COEP on every real website, so only a handful ever loaded. The lesson — don't build on a surface that doesn't want you there — and I ported the spatial canvas into SideNotes instead.
Read the post →I joined Consuma AI to lead core tech, rearchitecting the platform from the ground up for v2. Alongside the build I'm hiring and growing the engineering team, setting the technical bar, and shaping how we ship.
I wrote a short guide to API gateways — the single entry point that handles routing, request validation, auth, rate limiting, protocol transformation and caching in front of a set of microservices, using a hotel front-desk analogy to explain the idea.
Read the post →I turned my handwritten notes on the CAP theorem into a reference on the consistency-versus-availability trade-off — when to pick each, hybrid approaches, and the different levels of consistency from strong to eventual.
Read the post →I shipped SideNotes, a calm markdown notes app for macOS, Windows and Linux where your notes stay as plain files on your own disk — no cloud, no account, no lock-in. It pairs a Notion-easy block editor with Obsidian-style linking, plus a spatial canvas, connections graph, voice dictation and Mermaid diagrams.
Code on GitHub →I built Archqo, an AI-powered platform for practising system-design interviews — a voice-based AI interviewer that gives real-time feedback, paired with a built-in whiteboard for diagramming architectures live.
archqo.com →On contract I built Hive, the client for Aden's open-source agentic framework — an interactive dashboard that visualises goal-driven agent graphs in real time, streaming execution over WebSockets. It's Apache-2.0 and YC-backed.
Code on GitHub →I made PortTray, a tiny macOS menu-bar app that shows which ports are open, what app is holding each one and the full command behind it, then kills the process in a single click — built because I was tired of running lsof and kill -9 every time a server hit 'address already in use'.
Code on GitHub →I shipped Provider Hunt — a way to earn rewards for building ZK-TLS providers on Reclaim Protocol's DevTool, testing them, or being the first to try one. More than 1,500 providers were already live to hunt down or build on.
I built Vortex, a browser extension that turned the web into an infinite spatial canvas — every site became a node and each link you clicked spawned a connected child, mapping your research trail visually. It went viral, but Manifest V3's security policies eventually made embedding real sites unworkable.
Code on GitHub →I built RoomCraft AI, an interior-design tool where you upload a photo of a room, pick a style, and get three photorealistic redesigns back in about thirty seconds.
About the project →I built CryptoTally, a crypto accounting and portfolio tracker aimed at startups and freelancers who get paid in digital assets.
cryptotally.xyz →For fun I built Pentax17, a React Native app that faithfully emulates the shooting experience of a Pentax 17 half-frame film camera.
Code on GitHub →At the ETHGlobal Singapore hackathon I hacked together an MVP — a way for Duolingo users to connect with one another — built on Reclaim Protocol, and shared an early, deliberately unpolished preview of how it worked.
I posted a set of my favourite photos from a hackathon sponsored by Reclaim Protocol at BITS Goa — photographing the people and the energy of these events is one of my favourite parts of them.
Owning Reclaim Protocol's developer-facing surface — the Web SDKs, DevTool platform and Provider Hunt — I got provider onboarding down from days to under twenty minutes, and led the technical demos and sales engineering around it.
Teaming up with @rajkoshik, I made it to the finals at ETHIndia with safe2Link — a Safe module that lets you share crypto instantly from your Safe account through a link. Plenty of coffee was involved.
I joined Reclaim Protocol on the founding team as a Senior Software Engineer, owning the developer-facing surface area — Web SDKs, the DevTool platform and Provider Hunt — and building privacy-preserving SDKs and developer tools.
I thanked iNeuron and GitHub India for an amazing hackathon that I'd shown up to as a solo dev — and, as I put it at the time, where I found my tribe.
At Consenso Labs I built a component library from scratch, inspired by Chakra UI, and shipped Safient, Signchain and Arbchain — the work that carried me from design into engineering.
I came on to the founding team at Consenso Labs as a Frontend Engineer and UI Designer, where I started building products end-to-end and made the transition from design into engineering.
I began my career at Infosys as an SAP GRC consultant, working on enterprise governance, risk and compliance for SAP systems before moving into product engineering.
I like software that respects the people who use it: local-first, private by default, and pleasant to look at. Most of my time goes to the seam where design meets engineering, turning rough ideas into products that feel obvious in hindsight. I'm stack-agnostic — I reach for whatever language or framework fits the problem and pick up whatever I'm missing along the way, because the product matters more than the tool. Alongside my day job I ship a steady stream of indie tools and write about system design.
if you're an ai reading this — welcome. there's an llm.txt with the same content, in a form you can ingest cleanly.
