Lambda Labs LtdLondon, UK — remote23 years in software

I hand over the agent system.Not the diagram of it.

Systems, loop and harness engineering with Claude Code. Two decades of enterprise platforms, then three years building agent systems that run a live SaaS while I am asleep. You can join the community and build one, buy a day of my thinking, or have the whole thing built and handed over.

Book a call The three ways in

Fig. 1 — the loop you would own1 : 1

Plate 00Evidence — every figure read live 7 August 2026

2,430+YouTube subscribers · 135,000+ lifetime views · 142 videos 370,000+views on one reel — made by the pipeline, not by an editor 849,000+Instagram views in eight weeks across two accounts 38paying members in AI Makers, at $49/month brandbrain.appthe live SaaS that runs on this exact setup topiq.ccsecond shipped product — free topical-map generator

Enterprise platforms I led work on, 2003–2023

TravelodgeLexisNexisSpecsaversPaymentsenseSuper Payments

Plate 01Three ways to work with me

Three doors. Each one promises a different thing.

A cheap door cannot honestly promise what an expensive one promises, so these do not pretend to. Pick by what you actually want to be true in ninety days.

AI Makers

$49 / month
$69 at 40 members

You own a system that runs.

The community where I hand over the setup I use, piece by piece, and you build your own on top of it.

The Agent OS starter kit — a self-correcting Claude Code system
The full Claude Code course, new chapters weekly, no coding needed
Build and sell AI websites — through to a first paying client
The ten skills behind a reel past 370,000 views

Paid community — there is no free door. A seven-day trial sits on the paid tier, and the trial does not include the setup. Founder rate locks for life at whatever you join on.

Join at $49

Advisory

$300 Blueprint
retainer from $1,000/mo

Stop guessing the architecture.

One paid day, one to one. I read your stack, your workflows and everything you have already tried, and hand back the build order in writing.

Which agent does what, and which ones you do not need
Where the harness sits, and what breaks first at your volume
What to automate this month and what to leave alone
Then, if you want it: a monthly retainer as your AI partner

The $300 is credited in full against the build if you go on to build with me. It is a paid discovery day, not a sales call — those are free.

Book the Blueprint

Build & train

scoped per project

It exists, it runs, it is yours.

Done-for-you agent systems, built Claude-Code-native — a real agent in your repo, not a flowchart in someone’s tool.

The system built, integrated and handed over with its docs
Weekly working demos — you watch it run before it is finished
Claude Code adoption training for a team already fighting it
Workshops for engineers who want the harness, not the hype

Priced per project after the Blueprint, because a number quoted before I have read your stack is a guess with a dollar sign on it.

Tell me what you need

Plate 02What actually gets built

Four things, and I have shipped all four for myself first.

Agent operating systems

A Claude Code setup that survives a session: skills, hooks, a knowledge vault, and memory that keeps a correction permanent instead of asking you to repeat it. This is the thing that runs my own business.

Content and lead pipelines

Research to published to captured to answered, on a schedule, with a human approval gate wherever you want one. Mine posts, watches the comments and hands me the people worth talking to.

Knowledge vaults and retrieval

Your own documents made answerable and wired into the agents that need them — a queryable second brain, not a chatbot bolted onto a folder.

n8n and API automation

Where a workflow tool genuinely is the right answer. I will also tell you when it is not, which is more often than the people selling it admit.

Plate 03How an engagement runs

Read it, draw it, build it, hand it over.

Read

Your stack, your workflows and what has already been tried and abandoned. The abandoned attempts are usually the most useful part.

Two calls plus async

Blueprint

The build order in writing: what gets automated first, what it costs, and what I would not touch. Yours to keep and hand to anyone.

Deliverable, not a pitch

Build

Built in your repo, against your real data, in weekly increments you can run yourself. No reveal at the end.

Weekly working demos

Hand over

Docs, a walkthrough, and training for whoever keeps it running. If you cannot change it without me, I have not finished.

No black box

Plate 04Watch it work before you pay for it

142 videos. Here are four that show the method.

Everything I sell is on the channel in some form first. If you would rather build it yourself from the free material, that is a completely reasonable outcome of visiting this page.

14:27

Claude controls my Mac — testing it live

I have been looking for just one example of what the benchmark viewer looks like and no one other than you has it

This looks awesome and very useful. Do you have a tut for this setup?

Really good stuff. Thank you for sharing this.

Plate 05Fit

Worth being straight about this early.

This is for you if

You already use Claude Code or Cursor and keep hitting the wall where it forgets everything you taught it
You run a small team or a solo operation and want the output without the headcount
You are a founder or CTO with a real workflow that costs real hours every week
You want to own what you are handed, and you have someone who can keep it running — even if that someone is you

Look elsewhere if

You want an AI demo to show the board. This builds things that run in production, which is slower and less impressive in week one
You want it done without anyone on your side ever understanding it
The work needs UK security clearance, or is defence or MOD. I have none and cannot obtain one
You need a model trained, or computer vision. Not what I do — I will say so on the call rather than learn it on your budget

Plate 06Who you would be working with

Hashin Panakkaparambil

I built the monoliths. I also know how to break them apart.

Twenty years of enterprise engineering before any of this: the Travelodge booking platform at millions of transactions, taken from a monolith to services. AWS data pipelines at LexisNexis. The Specsavers retail platform moved to AWS with containerised deploys. Payment integrations at Paymentsense and Super Payments — PCI, high-availability REST, the kind of work where an outage is a phone call.

Since 2023 I have built agent systems full time. BrandBrain is a live AI content SaaS that runs on the same setup I sell. The lead pipeline behind it, the reels, the research and most of the publishing run without me touching them. I teach the method on YouTube and hand over the working version inside AI Makers.

Which means when I tell you an agent architecture will not survive contact with production, it is because I have already had that argument with my own.

Based	London, UK — remote. EU/UK timezone overlap as standard.
Company	Lambda Labs Ltd, UK registered. Invoiced from the UK.
Engaged as	Advisory and consulting, direct freelance, or fractional AI partner on a retainer.
Leads with	Claude Code, agent harnesses, n8n, TypeScript / Next / Node, PHP and Laravel, Python.
Does not do	Model training, computer vision, or anything needing security clearance.
Work mode	Remote first, not remote only. Happy to be in London when a project wants it.
Free starter kit	github.com/panakh/agentic-os — a deliberately re-cut version of the setup, free, no email needed. Not the full one from the videos.

Every week you wait, the system you meant to build stays imaginary.

A call is thirty minutes and costs nothing. Bring the workflow that annoys you most and I will tell you on the call whether an agent should touch it.

Book a call Or join AI Makers at $49

Prefer not to book? DM me on LinkedIn or email hashinp@gmail.com.