Skip to content
aadi.
Sheet 01/CoverX-ray lens — on autopilot

AdityaRajbhar.

Portrait of Aditya Rajbhar
Fig. 0 — The engineer

I build software that holds up under load — considered interfaces, honest APIs, and systems that scale without drama.

Drawn by
Aditya Rajbhar
Role
Software Engineer
Currently
BrainFog Technologies
Local time
India
2+
Years, professional
10+
Things shipped
6
Languages
Open to side projectsScale 1:1
Full-stack productDistributed systemsAI agentsVoice pipelinesInfrastructure
Sheet 02/ProfileGeneral arrangement

Pixels down
to queues.

I’m a software engineer from India who likes working the whole depth of the stack — from the pixels people touch down to the queues, workers and databases behind them.

At BrainFog Technologies I build native Android apps in Kotlin, full-stack web apps on Next.js, Python backends, and AI agents that automate real work. Before that I spent two years on the web platform for Landmark Aquatec’s water-treatment products: real-time dashboards over live plant telemetry, backed by FastAPI services on Google Cloud with Firestore.

On my own time I build agentic and infrastructure software: Upzy for uptime and attack-surface monitoring, Lookfar for self-hosted watcher agents, Avni, a voice-first assistant with a Kotlin Android app and a Rust core, and Keel, a SaaS starter kit.

I care about the decisions that don’t show up in a screenshot: where the boundaries are, what the data model actually says, what to leave out, and whether the next person can read it six months from now.

Sheet 03/Systems4 figures · 6 in the parts list

Drawn, built,
and still running.

Fig. 1—Upzy — probe, grade, verifyPan →
schedulerevery 60 sHTTPTCPDNSSSLyour-site.comaddress pinned at connect timeexternal surfacepassive scancerts · headers · DMARC · CVEsgraded a to fABCDFfinding + exact fixfor your stackverify fix ↻123
Platform2026In development

Upzy

Know it’s down before your customers do

Uptime monitoring and external attack-surface security for small teams, in one tool. Checks run every minute, passive scans grade each site A to F, and every finding comes with the exact fix for your stack and a Verify fix button.

  1. HTTP, TCP, DNS and SSL checks, heartbeats, and public status pages with custom domains
  2. Passive scans for certificates, headers, SPF/DMARC, subdomain takeover and CVEs ranked with CISA KEV and EPSS
  3. SSRF-safe probes that pin the resolved address at connect time; opt-in AI on your own key

Bill of materials

  • Bun
  • Turborepo
  • ElysiaJS
  • Next.js
  • PostgreSQL
  • Drizzle
  • Better Auth
Fig. 2—Lookfar — no change, no tokensPan →
sourcespages · feeds · reposfetchnormalisehashdiff?samestop · 0 tokenschangedmodelonly what mattersreportapprovals inboxexact payload shownactonly after a yespostgres job queueFOR UPDATE SKIP LOCKEDschedules123
AI agents2026In development

Lookfar

Agents that look far, so you don’t have to

Self-hosted watcher agents you can leave running. Describe what to watch in plain language and Lookfar checks pages, feeds and repos on a schedule, reports only what matters, and asks before it acts on your behalf. It grew out of Feedoku, my earlier RSS aggregator.

  1. No change, no tokens: a free fetch, normalise, hash and diff gate runs before any model is called
  2. Anything with side effects waits in an approvals inbox with the exact payload shown
  3. Postgres job queue on FOR UPDATE SKIP LOCKED, SSRF-guarded fetcher, hashed API tokens

Bill of materials

  • Go
  • SvelteKit
  • PostgreSQL
  • pgvector
  • Better Auth
  • Docker
Fig. 3—Avni — cheapest tier that can answerPan →
speechSTTrust coretiered routerT0 · on-device rulesT1 · fast LLMT2 · strong LLMT3 · background agent jobsslower + costlier ↓TTS≈1.0 sp50 · stt done → first replyMCP tools & skillsswappable STT · TTS · LLM123
Voice AI2026In development

Avni

A voice layer for everything you own

A speech-first personal assistant that answers in about a second, acts on your devices and services, and asks before anything risky. An Android app talks to a single Rust core that routes each request to the cheapest model tier that can handle it.

  1. Measured over 117 real turns: p50 about 1.0 s from speech-to-text done to first reply text
  2. Tiered router: on-device rules, then a fast LLM, then a strong LLM, then background agent jobs
  3. Every STT, TTS and LLM is a swappable provider; tools and skills plug in over MCP

Bill of materials

  • Rust
  • Kotlin
  • Android
  • MCP
  • SQLite
  • Astro
Visit livePrivate repo
Fig. 4—Keel — wired once, ready on day onePan →
keel/apps/apiElysiaJS on Bunapps/webNext.js App Routerpackages/dbPrisma 8packages/authBetter Auth · orgspackages/emailReact Emailpackages/uishared kitpackages/envzod-validated✓ typed api · auth · email · ui · env — before the first feature123
Infrastructure2026Open source

Keel

The backbone every SaaS ends up rebuilding

An opinionated Bun + Turborepo monorepo that wires up the boring-but-critical parts once — a typed API, auth with organisations and admin, transactional email, a shared component kit and a validated environment — so a real product can start on day one.

  1. ElysiaJS API on Bun, Next.js App Router web app
  2. Prisma 8 data layer, Better Auth with org + admin plugins
  3. React Email via Resend, zod-validated env, oxlint / oxfmt

Bill of materials

  • Bun
  • Turborepo
  • Next.js
  • ElysiaJS
  • Prisma
  • Better Auth
Sheet 04/Revisions31 months in production

Revision history.

  1. B

    Sep 2026 → Present

    Current

    Full Stack Developer

    BrainFog TechnologiesMumbai, India

    Building across the whole product surface: native Android apps in Kotlin, full-stack web apps on Next.js and Prisma, Python backends, and AI agents and automation with LangChain, LangGraph and MCP. It’s also where I picked up Kotlin.

    • Kotlin
    • Android
    • Next.js
    • Prisma
    • Python
    • LangChain
    • LangGraph
    • MCP
  2. A

    May 2024 → Sep 2026

    Superseded

    Software Developer

    Landmark AquatecMumbai, India

    Built the web platform behind the company’s water- and sewage-treatment products: real-time dashboards over live plant telemetry (sensors, pumps, valves), status and service reporting, and role-based administration for sites and customers. FastAPI services on Google Cloud with Firestore, behind a React / Next.js front end.

    • Next.js
    • React
    • TypeScript
    • FastAPI
    • Python
    • Google Cloud
    • Firestore
Sheet 05/StackExploded view · 44 parts

The stack, in section.

    • React
    • Next.js
    • Tailwind CSS
    • shadcn/ui
    • SvelteKit
    • Astro
    • Ant Design
    • React Query
    • Node.js
    • NestJS
    • ElysiaJS
    • Better Auth
    • FastAPI
    • Express
    • Spring Boot
    • GraphQL
    • MCP
    • LangGraph
    • Agents & tool use
    • OpenAI API
    • Google Gemini
    • RAG pipelines
    • PostgreSQL
    • Firestore
    • MongoDB
    • Redis
    • Prisma
    • Drizzle
    • pgvector
    • SQLite
    • Kafka
    • Google Cloud
    • AWS
    • Docker
    • Turborepo
    • Bun
    • Vercel
    • Netlify
    • TypeScript
    • JavaScript
    • Python
    • Go
    • Rust
    • Kotlin

Broadcast / YouTube @codebyaadi

New shorts
every week.

Quick builds, real debugging sessions and the occasional rant about tabs vs. spaces — no scripts, no filler.

YouTubeWatch on YouTube
Sheet 06/Notes4 general notes

General notes.

N01What does Aditya Rajbhar do?

Aditya is a software engineer from India. He is a Full Stack Developer at BrainFog Technologies, building Android apps in Kotlin, Next.js web apps, Python backends and AI agents. Outside work he builds Upzy (uptime and security monitoring), Lookfar (self-hosted watcher agents), Avni (a voice assistant) and Keel (a SaaS starter kit).

N02What is his technical stack?

Day to day: TypeScript with React and Next.js on the web and Kotlin on Android; Node.js, ElysiaJS, NestJS, FastAPI, Go and Rust on the back end, with SvelteKit where it fits; PostgreSQL, Firestore, Redis and Prisma/Drizzle for data; LangGraph and MCP for agents; Docker, Turborepo and Google Cloud for infrastructure.

N03Is he open to new opportunities?

Aditya works full-time at BrainFog Technologies and is not looking for a new role. He is open to side projects and collaborations involving full-stack product work, distributed systems, or AI agents and tooling.

N04How do you get in touch?

The fastest way is a direct message on X (@codebyaadi). He’s also on GitHub and LinkedIn as codebyaadi.

Sheet 07/ContactOpen to side projects

Got a system
worth drawing up?

I work full-time at BrainFog, so I’m not looking for a new role — but I’m always up for a good problem: a side project to build, a system to untangle, or an idea worth pairing on. A direct message on X with a real question is the fastest way in. Cold sales pitches get ignored.

Message me on X