Blimp — AI agents, apps and automation, built and run for you.
AI agents, apps and automation — built and run for you.
Freelance AI engineer · Kuala Lumpur · GMT+8
I design, build and run AI agents, SaaS products and automation, end to end. You tell me what you need; I ship it, then keep it running.
Or build your own AI agent yourself — coming soon.
Services
Three things I build
Designed, built, hosted and supported end to end — pick a lane.
Your own AI agent on WhatsApp, Telegram or Discord.
Tuned to how you actually work. Answers around the clock, with real long-term memory — walled off from everyone else's.
- Tuned to how you actually work
- Answers around the clock, with real long-term memory
- Budget-hosted; your data walled off from everyone else's
Built for real people
Full products, end to end.
From data model to billing — designed, built and run.
- Full products — data model to billing
- Cited, grounded answers
- Next.js · Supabase · Vercel · CI/CD
Annota (annota.my) · live
Self-running multi-agent systems.
Blimp itself runs on a three-agent studio sharing one brain.
- Multiple agents working together
- Runs itself, watched for problems
- One shared memory across agents
Built solo · ~6 weeks to self-running
Your own AI agent on WhatsApp, Telegram or Discord.
Tuned to how you actually work. Answers around the clock, with real long-term memory — walled off from everyone else's.
- Tuned to how you actually work
- Answers around the clock, with real long-term memory
- Budget-hosted; your data walled off from everyone else's
Built for real people
Full products, end to end.
From data model to billing — designed, built and run.
- Full products — data model to billing
- Cited, grounded answers
- Next.js · Supabase · Vercel · CI/CD
Annota (annota.my) · live
Self-running multi-agent systems.
Blimp itself runs on a three-agent studio sharing one brain.
- Multiple agents working together
- Runs itself, watched for problems
- One shared memory across agents
Built solo · ~6 weeks to self-running
The stack behind it
AI & agents
- LLM orchestration (Claude, open-weights)
- Multi-agent systems (MCP, A2A)
- Agent deployment & hosting (Telegram · WhatsApp · Discord)
- Self-improving agents
- Prompt engineering
- RAG & citation grounding
- Evaluation harnesses
- Deterministic-core design
SaaS & platform
- TypeScript / Next.js
- Python
- Supabase / Postgres
- Vercel
- AWS
- CI/CD (GitHub Actions)
- Stripe / merchant-of-record billing
How it works
Two calls, then I keep it running
We talk, I build, and it stays looked after.
Understand
One call to understand what you need — workflow, edges, data.
Set it up
A second call to set it up — wired to your channel, tuned to you.
Maintain
Then it stays live — tuning, monitoring, support.
The work
What people say
Real words from people I've built agents for — verbatim, never invented.
The level of support I get when I have an issue is great.
It's very convenient that the agent is just one WhatsApp message away, wherever I am.
The agent remembers things I told it months ago.
Way cheaper than the ridiculous monthly AI subscriptions nowadays.
Solid dungeon master for my D&D campaign!
Live demo
Describe it. Watch me design it.
Tell me what you want built — I'll design the whole system live. Real AI, right here.
The studio, live
A studio that runs itself.
Three AI agents — Duke, Hermes, Frank — one shared brain, running the studio around the clock.
Nobody's at the wheel. It just keeps running.
- Heartbeathealth check, on a timer
- Overnightdrafts + self-verifies
- Dreamsreflects, captures lessons
- Backupstwo-tier, off-box
- Weekly evalsbehaviour regression suite
Activity
Duke logged a decision to the shared brain
The real shape of it — three agents, one shared memory, keeping the lights on 24/7.
Selected work
Things I've shipped.
Real Blimp products, built end to end. Hover to explore.
Annota
AI employment-law guidance
Cited guidance over ~11,000 real Industrial Court awards, with a deterministic citation gate — the model never invents a case. The LLM explains; deterministic code decides what's binding.
- Next.js
- Supabase
- RAG
- Citation gate
- Stripe
IR Assist
A caseload workspace for practitioners
A multi-tenant workspace over the same legal corpus — full-text search, cited answers, and billing, built end to end.
- Multi-tenant SaaS
- Full-text search
- Postgres
- Stripe
The Blimp studio
A studio that runs itself
Three AI agents on one shared brain, wired over A2A with evidence-gated task completion — self-running, watched for problems. It builds and operates everything you see here.
- Multi-agent
- MCP
- A2A
- Self-improving
Who's behind it
Logan Chandrasegaran
Founder · Freelance AI engineer
Kuala Lumpur · GMT+8
I'm Logan. I run Blimp solo — AI agents on any channel, SaaS products, automation, built and run end to end. I ship, then make it better.
Start a project
Tell me what you want to build
A rough sentence is enough. No obligation — a real reply from a real engineer.
Contact
Have something in mind? Let's build it.
A rough sentence is enough — you'll get a real, considered reply from the engineer who'd actually build it. Kuala Lumpur · GMT+8.
Want to build your own agent?
Self-serve is opening in batches — leave your email.