Hi, I am Mayank. I build machines that run themselves, and everything you see here is live.
Four Instagram accounts that publish without me. An ad engine that spends real money daily. A study system that tells me what to revise next and admits when the plan doesn't fit.
Every number on this page is read live from the system it describes. Where a machine can't reach its source it says so, rather than showing you something invented.
4 of 5 sources reporting · read at 01:56 UTC
Each of these is a system I designed, built and still operate. They're described as machines because that is honestly what they are — schedulers, ledgers, retries and telemetry.
Autonomous publishing across four Instagram accounts
Content is generated, rendered to image, deduplicated against a permanent ledger, and published on a schedule — with no human in the loop. Three separate duplicate incidents taught me not to trust any single guard, so there are now four independent layers, and a lost ledger write raises rather than failing quietly.
AMNICENT's client roster
Five real clients across tutoring, HR-tech, a marketplace app, local accommodation and student research. Paid media, websites and content systems. Every cable on this panel is live work, not a sample.
Open →Paid media — Meta and Google
A two-country campaign built against the Meta Marketing API directly — no SDK, plain HTTP. Singapore's advertiser-compliance requirements alone cost several failed attempts before the exact working shape was found; it is now encoded in config so it never has to be rediscovered.
LinkedIn publishing, with a human approval gate
The one system here that deliberately will not act alone. It drafts, then asks on Telegram, and publishes only on an explicit yes. No reply means no post — a missed notification costs a day of cadence rather than putting something live unreviewed. The automation is the courier; the human is still the approver.
Life OS — study, focus and life tracking
A tracker built around one date: the CFA Level I exam. It plans what to study next from exam weight, measured mastery and staleness, and it reports honestly when the plan does not fit — from a standing start, the full curriculum does not compress into the days remaining, and it says so instead of emitting a fantasy schedule.
Everything that gets built once and stays built
Print-ready workbooks pitched to international schools, an AP Calculus revision engine that typesets real mathematics and audits its own page overflow, and a video pipeline that writes finance explainers by hand on screen with a synthesised voiceover.
Through AMNICENT. Each cable is a real engagement — paid media, websites and content systems. Not samples, and not case studies with invented percentages attached.
What you're reading is the content and telemetry layer of a build that becomes a walkable 3D machine room — camera on rails by default, free-roam if you want the controls, and one door into a photoreal capture of the desk all of this was actually built at. It ships in layers, so it is never half-finished in public.