less software.
fewer hours.
same output.
i help founder-led service businesses cut manual work out of their operations — without adding another tool nobody maintains.
sometimes that means an existing product you haven't found yet. sometimes an open-source project that replaces a subscription. sometimes software built specifically for how you work. i do all three, so my recommendation isn't decided by what i happen to sell.
and i publish what breaks, not just what ships — because the interesting part of automation is never the happy path.
fifteen minutes, no slides. if i can't help, i'll tell you who can.
costed at $50 an hour — the blended cost of the people currently doing it. not every line on a map like this gets removed; some things shouldn't be automated, and i'll say so.
for founder-led service businesses doing $1m–$10m a year, where growth is outrunning the team's capacity to deliver.
i map where your team's hours actually go, remove the highest-cost manual work, and give you back the equivalent of a part-time hire — usually within six weeks.
your process lives in people's heads, not in systems.
every job runs through somebody's judgement. quality changes depending on who picks it up, deadlines slip whenever someone is away, and the only way to take on more work is for everyone to do more work. that's a business with a hard ceiling — and one that's difficult to sell, because the value walks out at six o'clock.
your best people spend their week on work a system should do.
copying data between tools. chasing documents. compiling the same report every friday. sending the follow-up that should have sent itself. your most capable people didn't join to do this, they know it's beneath them, and the ones with options will eventually act on that.
you already tried automating, and it broke.
someone wired your tools together. it worked in the demo. then an edge case appeared, something failed silently, and nobody could tell you why. now you have an abandoned workflow, a subscription you still pay for, and a team that has quietly gone back to doing it by hand. this is the most common situation i'm called into — and it's fixable.
your software bill grows every quarter and nobody owns the stack.
fourteen subscriptions. four of them overlap. two are barely used. nobody has reviewed the list in eighteen months because it's nobody's job. frequently the first thing i do is reduce that bill — which isn't a service most people offering ai will ever suggest to you.
map. remove. maintain.
i sit with you and your team and find where the hours actually go. you get a written map of every manual process, costed in hours and money, ranked by what's worth fixing first.
then the cheapest thing that works. an existing product if one fits. an open-source deployment if that's better value. custom software when the workflow is genuinely yours. built properly either way: tested, monitored, documented.
systems drift. models change, apis change, your business changes. i watch the systems i build, fix them when they break, and send you a monthly note on what ran, what broke, and what it saved you.
i'll talk you out of hiring me if that's the right answer.
most ai implementers can only sell you the thing they know how to build. if they build no-code workflows, your problem needs a no-code workflow. if they write software, your problem needs software.
i can do all three, which means i have no reason to steer you. roughly a third of the audits i run end with a recommendation to buy an existing product or deploy something open-source, and i'll help you do that. it's a smaller invoice and a much better outcome.
the order matters. most people in this market start at row three, because it's the only row they can deliver.
from production systems to practical operations.
i'm a software engineer. i spend my days building systems in fintech — an industry where a process failing quietly is expensive, and where nobody is impressed by a demo.
that's where i learned the thing that separates automation that survives from automation that doesn't: the interesting part is never the happy path. it's what the system does when a document is malformed, an api is down, or the model returns something confidently wrong.
today a yc-backed us startup runs systems i built, handling live customer requests. i contribute to the open-source ai frameworks i work with, which is how i stay current and how the cto of one of them came to recommend me.
i bring the same standard to businesses far smaller than a bank — because a fifty-person firm can afford a broken process even less.
a yc-backed us startup runs agent systems i built against live customer traffic — artifact.engineer.
the cto of mastra — the ai agent framework — referred me personally, after roughly ten merged pull requests.
four tools i built and run in public. working software you can open and try, rather than a description of working software.
a week of my time to map where your hours go, and a straight answer on what to buy, what to deploy and what to build. refunded if i can't find you five hours a week.
send me the workflow that broke. you get a written explanation of why it failed, what it would cost to make it reliable, and a straight answer on whether it's worth saving. credited against the fix.
the right existing product or open-source project, chosen, deployed, integrated and taught to your team.
software built for how your business actually works, with testing, error handling, monitoring and full handover.
i watch what i built, fix it when it breaks, improve it monthly, and send you a note on what it saved.
if your problem doesn't fit one of these, tell me about it and you'll have a fixed number in writing within 48 hours.
all prices in usd.
fixed fees. never hourly.
“you're one person.”
yes. that means the person you speak to is the person who builds it — no account manager, no handoff to a junior. it also means i take on one build and a handful of ongoing clients at a time, and i say so publicly. you get everything: code, documentation, credentials. if i disappeared tomorrow, another engineer could pick it up on monday.
“you're not in our country.”
i work from india, and my afternoons are your mornings — i hold fixed hours for uk and european clients and answer within them. everything runs in your own cloud accounts, not mine. we'll speak on video, often.
“we've been burned before.”
reasonably. that's why the first step is a fixed-fee audit with a refund if i can't find you at least five hours a week worth reclaiming — and why every system i build ships with monitoring and alerts, so you find out something broke from me rather than from a client.
“is our data safe?”
an nda before anything moves — i'll send it before you ask for it. everything runs in your own infrastructure, under your accounts and your api keys, so your data never sits on my machines. i take the minimum access that does the job and hand it back at the end. anything else you need in writing, we cover on the call.
i build my own tools in the open. it's where the claims above can be checked by anyone who wants to.
deploy personal ai agents to your own infrastructure with one click.
ai agent that learns your voice and creates x content ideas from top creators.
personal ai teacher with socratic questioning and spaced repetition.
database and memory management system for ai agents with drizzle and libsql.
one real system. every week.
each week i take one manual process out of a real business and show you exactly how it was replaced — the tool or the code, what it cost, and what broke along the way. no predictions, no hype, nothing you can't use on monday.
free. weekly. unsubscribe whenever you like.