A support agent that
handles the easy stuff.
Resolves routine storefront requests using order and policy context. Money decisions and sensitive cases go to a person.
ever has to touch.
I turn messy support queues, returns and shipping decisions into systems your team trusts. Built for production, not for the demo.
Operator turned engineer. Leading automation at Nysonian, co-founder of ZavrAI.
Less theatre.
More things that work.
automations running in production
less manual returns work, estimated
orders analysed for one shipping decision
The opinion I keep coming back to
Start with the task your team repeats every single day. Make it dependable. Then earn the complexity.
I started on the operations floor. I find where the work actually breaks, on real orders and real handoffs, before I write a line of it.
The order, the policy, the tool. Retrieve what is needed when it is needed. A bounded job for the agent, and everything deterministic stays deterministic.
Retries, monitoring, the exception nobody mentioned. I ship it, then keep it alive alongside the people who depend on it every day.
Built and maintained alongside the people who use them.
Resolves routine storefront requests using order and policy context. Money decisions and sensitive cases go to a person.
Connects return labels, warehouse scans, customer updates and refund preparation in one tracked lifecycle.
Replaces disconnected spreadsheets with a shared view of backorders, fulfillment, containers, returns and invoices.
Adds label-first, multi-order picking to ShipHero: box planning, label purchasing, aisle-ordered pick sheets and shipment sync.
Bilal, pronounced Bee-Laal
The human in the loop
The floor. The orders. The handoffs. I knew the friction before I ever wrote the code to remove it.
Today I lead automation at Nysonian and co-founded ZavrAI. I build AI agents, context pipelines and operational systems for a US ecommerce business, with inventory kept in sync across six countries.
My honest opinion after shipping these: earn the complexity. Start with something dependable and boring, then grow it once people trust it.
ECS. Ran the floor. Learned exactly where work breaks.
Nysonian. Mapped the manual processes worth automating.
Nysonian. Built the automation program from scratch.
Nysonian. Promoted in six months. Own delivery and the team.
The tools follow
the problem.
Claude API, Tool calling, RAG, Ollama, turbovec
TypeScript, Python, n8n, PostgreSQL, Shopify, ShipHero
Before you ask
Short, honest answers for recruiters, operators and founders sizing up whether to work together.
I build AI agents and automation systems that take repetitive operational work off your team: customer support triage, returns and RMA lifecycles, and order or shipping decisions. I connect the tools, give the agent the right context, and keep it running in production, not just in a demo.
No. Your first agent should be boring. Start with one task your team repeats every day, make it dependable, and earn the complexity from there. Ambitious agents nobody trusts get switched off within a month.
Yes. A well-built support agent handles routine, low-risk requests like order status and closes them quickly. Money-related and sensitive conversations escalate to a person by design. The goal is a shorter queue, not a wall between customers and your team.
Claude API and tool calling for agents, n8n and Python for workflows, PostgreSQL for data, and integrations with Shopify, ShipHero and carrier APIs. I also work with RAG, Ollama and TypeScript. I pick the tool that fits the problem rather than forcing the problem into a tool.
A focused first automation is usually a few weeks from scoping to something live in production. The unglamorous part, retries, monitoring and the edge cases nobody mentioned, is part of the work, not an afterthought bolted on later.
I lead automation at Nysonian and co-founded ZavrAI, and I take on selected projects and roles. I work remotely with teams anywhere. The fastest way to check fit is a short discovery call.
Your next chapter
A bottleneck to untangle. A returns mess. A support queue that never ends. Let's scope it and see what a dependable system would actually save you.
Hi, thanks for scrolling this far.
To me, good automation is quiet. It just works, every day, without anyone thinking about it.
If something here stuck with you, say hello.