Systems Architect at Aguru AI
Your SaaS has an AI strategy. It lacks the infra to make AI do real work.
Build it the way OpenAI, Replit, and Cursor do.
At Aguru, I shipped the durable execution engine in 4 weeks, after two failed attempts over 9 months. It now runs AutoCert in production — a highly sophisticated, long-running, customer-facing AI workflow that executes end-to-end for weeks, unattended.
Same playbook for your SaaS: the AI durable execution layer (Temporal for unmatched reliability, Kubernetes for unmatched scalability from day one) in your stack, with the first customer-facing AI workflow shipped on it — reliably, end-to-end. Your product does the work, your customers get the outcome. Your team ships every workflow after that, without me.
Real AI work
Standard infra runs features. AI doing real work needs a reliable execution layer.
Signs your AI needs a reliable execution layer:
Long-lived
Runs for days, weeks, or months — a process with a lifecycle, not a request and a response.
Cross-system
Coordinates work across your own services, third-party APIs, and external data sources. The work lives between them, not in one place.
Multi-stakeholder, human-in-the-loop
Multiple parties are involved. Models decide what they can; humans decide what they can't.
Reliable at scale
Survives deploys, restarts, and infrastructure failures. Loses no work, drops no state, leaves nothing stuck.
Auditable
Enterprise-grade observability and governance — full visibility into who did what, when, and why.
Not every AI workload needs this. Dashboards, single LLM calls, and analytics are features — they run fine on standard infra.
Case study
Before AutoCert, every renewal ran the same way: an agency operator chased the Gas Safe engineer, booked the tenant, followed up over weeks, collected the certificate, filed it with the regulator, and absorbed every failure when a party went silent. A lot of scattered manual work across Alto. Alto could track the work. It couldn't do it.
We built AutoCert inside Alto — an AI agent that runs the entire renewal end-to-end. But the agent isn't the trick. The durable execution layer underneath is. Temporal handles the orchestration: long-running, fail-proof, automatically retried, and fully observable. Kubernetes scales it across thousands of concurrent renewals. The workflow uses that foundation to coordinate landlord, tenant, supplier, and regulator over weeks — with human handoffs when the model's confidence drops. Without the layer underneath, none of it ships.
Today Alto's agencies don't run the renewal — they receive the certificate. The product surface is the same; the layer underneath turned Alto from system of record into system of work.
How it works
An engagement built around shipping one customer-facing AI workflow into your platform — and leaving the durable execution layer with your team.
Step 01
~1 week
Together we identify the single AI workflow that delivers the biggest ROI to your users.
You leave with a written scope: the workflow, the stakeholders involved, the systems it touches, and the success metric we'll measure against.
You can walk away after this if it's not the right fit — no commitment to step 02.
Step 02
4–6 weeks
Embedded with your team. I architect and ship the workflow alongside the AI durable execution layer (Temporal + Kubernetes) into your stack, with 100% test coverage in CI/CD.
Mid-engagement, the workflow runs in shadow mode against real customer data. End of engagement, it ships.
Step 03
~1 week
Pairing with your engineers until they own it. By the end: your team can ship the next workflow without me.
What stays: the execution layer, the workflow runbook, the testing framework, and the architectural decisions written down.
Founders, CTOs, and product leaders I've worked alongside on production systems.
I've worked with a lot of engineers over the years, but few have combined deep technical ownership with the kind of composure under pressure that Ahmed showed on what became one of the most consequential builds in Aguru's early history.
The execution engine refactor was critical. The stakes were high — it sat at the core of our production infrastructure, with a hard deadline, no real fallback, and significant team changes happening in parallel.
Ahmed took full ownership: not just of the code, but of the outcome. That meant 12-hour days, weekend work, and staying locked in when the pressure was at its peak.
What sets Ahmed apart is that he thinks like a founder, not just an engineer. Having built and led his own startup, he brings architectural judgement and a sense of commercial stakes that you rarely find.
He didn't just ship — he shipped something we could all stand behind.The engine is now live in production and we're on schedule.
For any founder or engineering leader considering working with Ahmed: this is someone who shows up when it matters most. I am delighted to be working with Ahmed and look forward to building Aguru into a category leader in AI infrastructure.
Ahmed is one of those rare engineers who is both deeply technical and genuinely cares about the business. He's highly proactive and never waits to be told what to do.
He led the end-to-end re-architecture of our workflow engine under very tight deadlines. He has a strong attention to detail and a deep understanding of the product.
He's also remarkably responsive — he picks up context quickly, asks the right questions, and stays accountable through to the end. On top of that, he has a strong entrepreneurial mindset: every change is evaluated in terms of both its cost and the value it unlocks.
Your work has been a game-changer for EMURGO Fintech.
With a proven track record of excellence in extending our platform's capabilities and now transitioning to enhance our libraries, your expertise is a rare and valuable virtue.
Your adaptability and technical prowess ensure that we can confidently rely on your skills across the full spectrum of our software development needs.
Ahmed is a committed and resourceful engineer, always looking to solve problems effectively, learn new things and reach a successful conclusion.
He picked up the challenge of integrating durable execution into our platform: a technically challenging distributed systems problem that had defeated others.
He presented his approach, took feedback and revised the plan, researching areas where he didn't have experience to come back with solid recommendations and an approach that the entire team accepted.
He worked long hours to get it implemented within a challenging deadline, setting the direction and providing technical leadership.
Brought in during a tense phase, Ahmed identified bottlenecks and proposed solutions the team gladly adopted.
He communicates boldly and clearly, and immerses himself completely, end to end — he lives inside the product he develops.
You will do a great service to yourself, your team, and your product if you manage to attract Ahmed.
Ahmed has been a key pillar in our team and glad he's part of it.
He has a great work ethic and will go the extra mile to see the work done even in challenging circumstances.
He can work independently with minimal supervision and communication with him is always straightforward.