Open to backend / SDE roles — remote or relocation
I'm Kartikeyan Kidambi — a Computer Science final-year at KL University who writes Java and Spring Boot for a living and ships it to AWS. Nine systems are live right now: a commerce platform with role-based auth, a reservation system where double-booking is impossible, and an Android price comparator that keeps every byte on-device. I build with AI in the loop — agents, tool calling and spec-driven delivery.
Compare. Choose. Save.
Blinkit, Zepto and Instamart rarely agree on what a packet of atta costs, and checking each by hand is tedious enough that nobody does it. Switch does the checking — six stores, one search, priced for your own pincode. It is on real phones, in real kitchens, being used to decide real purchases.
Matching products are ranked by price per kilo or litre, so a bigger pack cannot masquerade as a better deal, and sold-out items can never win. Sign-in happens on the store's own page — no account, no ads, no analytics, and nothing identifying the user ever leaves the device.
Real live comparison · Android & Web · free, zero accounts
Food intelligence with an uncompromising trust layer.
Point a phone at the barcode on packaged food and NutriScan returns what is actually on the label — nutrition, allergens and a plain-language ingredient decoder — scored out of ten with the reasons shown. Photograph a meal or a restaurant menu and it estimates the plate against authentic nutritional references, letting you correct dishes and portions before anything is counted.
The part that matters most is the trust layer. Barcodes are mathematically validated before any lookup, data comes from verified label records, and anything beyond confirmed label data is marked as an estimate — never presented as fact. Non-food items are refused outright, and your privacy is respected with zero tracking.
Real screen from live app · Instant barcode scanning
Alongside our two flagships, Switch and NutriScan, every project below is deployed and reachable right now — no screenshots standing in for software. Open any card for the architecture, or jump straight to the running app.
A complete Indian storefront with three access tiers, a simulated Razorpay gateway that verifies payment signatures, and an admin dashboard reporting sales, order status and low-stock alerts.
Distinguishes alcoholic from non-alcoholic fatty liver disease from patient parameters, then shows which inputs drove the result and exports a PDF report. A prediction a clinician can argue with, not an opaque score.
Seat-level booking with real-time availability validation that makes double-booking structurally impossible, plus a coupon engine handling flat and percentage offers against dynamic fares.
Play a full, rule-legal game of chess with your hands in front of a webcam. The ML inference runs entirely in the browser — there is no server to send your camera feed to.
Files stream device-to-device over a WebRTC data channel. No upload, no bucket, no retention — and therefore no storage bill and nothing to breach.
A Spring Boot WebSocket server broadcasting messages in real time, containerised with Docker and deployed to the cloud behind a glassmorphic client.
A lightweight Excalidraw: multi-board canvas with shapes, sticky notes, zoom and full undo/redo history, built as a TypeScript pnpm monorepo.
Students, faculty, courses, attendance and grade points under one layered Spring Boot backend — controller, service, repository, entity — containerised and deployed.
Depth in the Java backend stack, working fluency everywhere else, and an honest line between the two.
Where I'm deepest — daily driver.
Deployed, not just studied.
Enough to ship the whole product myself.
How I build, and what I build into products.
The everyday workbench.
The part that outlives any framework.
An always-on autonomous agent on a cloud VM, reachable from Telegram, WhatsApp and Gmail. I built it because I wanted to know what it takes to keep an LLM agent running unattended — auth, webhooks, tool routing, failover — rather than only to read about it.
Webhook listeners on Telegram and WhatsApp; instructions arrive as ordinary messages.
A key-and-model chain retries a second provider when the first rejects, so a bad key never becomes an outage.
The chat bubble in the corner is the same agent. Ask it about my work — it's live, not scripted.
Education
Koneru Lakshmaiah Education Foundation (KL University)
CGPA 8.67 / 10Narayana Junior College
Sri Venkateswara Bala Kuteer
I'm looking for a backend or SDE role — open to remote teams anywhere, and happy to relocate. If you're hiring, or just want to know how one of these systems works, email is the fastest route and I reply the same day.