Moiz Ali
Tetovo Available for projects

AI agents & automationOnline · Tetovo

I build agents that do the work.

  1. 14:02:11.004 trigger: request · "Reconcile March invoices"
  2. 14:02:11.630 plan: 3 steps · 2 tools · 1 check
  3. 14:02:13.902 act: sheets → ledger · 41 rows matched
  4. 14:02:19.247 ship: verified · report sent to #finance
  5. 14:02:19.250 done: 8.2s · 0 errors · next run: yours ↓

I design, build and run AI agents and automations for teams that are done doing the same work twice.

Start a project Try it on your task 
02 — Your turnDescribe a task · see the system I would buildTry it

That was one of mine. Now yours.

What should run itself?

Try
Waiting for your taskWaiting for your taskThe system appears here—

This is how I start every project: map the work, decide where a model helps, where code is safer, and where a person stays in the loop.

See real runsBuild this with me

03 — How I work

03 / 08

Agents that fail loudly. Automations with an off switch. Systems your team actually trusts.

04 — Selected runs

03 runs03
RUN #003Cold outreach agentResearched cold emails where every claim has a sourceClaude · Python · FastAPI · Postgres · RedisBuilding2026RUN #002ConcordA supervisor that keeps multi-agent runs alive through crashesClaude · Python · Redis · OpenTelemetryBuilding2026RUN #001World Cup 2026 predictorRanks all 48 teams by their chance of winning the 2026 World CupPython · scikit-learn · XGBoost · SQLiteCompleted2026
A drafted email to PostHog next to the source URL and exact quote behind each claim, from a real run on the deployed system
RUN #003BuildingCold outreach agentResearched cold emails where every claim has a sourceRUN #002BuildingConcordRUN #001CompletedWorld Cup 2026 predictor
All runs 3

05 — Process

Trigger → Plan → Act → Verify → Ship
  1. Trigger01

    Trigger

    Something happens: an email lands, a form is sent, a schedule fires. I find the moment where the work really starts.

  2. Model02

    Plan

    The agent decides what needs doing and which tools it may use. Limits come first, so it never wanders.

  3. Tool03

    Act

    It does the work in your real tools (CRM, sheets, inbox) with the same access a person on your team would have.

  4. Code04

    Verify

    Every result is checked by code or a second model. When it is unsure, it stops and asks a person instead of guessing.

  5. Tool05

    Ship

    The result lands where your team already works, and every run is logged so it can be replayed later.

06 — Talk to my agentDemo replay · reads only my case studies

Ask it about my work. It will show you.

It answers from my case studies, shows every tool call it makes, and can take you straight to what you are looking for.

Agent · session 3f2aActing on the page
show me your best project
search_projects(sort: "impact")142ms · 4 results
ui.scroll_to("work")done
ui.highlight_project("[slug]")running
ui.open_project("[slug]")queued

This is RUN #011, [Invoice extraction pipeline]. I have highlighted it on the page and I am opening the full run report next.

07 — Toolbox

16 tools · hover to see where they ran

Models

Backend

Data

Ops

08 — ContactStart a run

Start a run.

Describe the work in your own words. I read every message myself and reply from

Book a call · [booking link]