Anotidaishe Guri
Remote · Harare, UTC+2 · available now
AI builder

I build working software with AI, fast.

Lending platforms, anti-counterfeit systems, health-risk models and the internal tools businesses run on, built with Claude Code, Codex and Cursor. I turn a problem described in plain English into a working product.

Farmers are buying fake seed. Can they check a bag with their phone?

plan · QR code per bag → verify → flag copies → live map

build · Python, FastAPI, Playwright tests, written with AI coding tools

Farmer scans a bag, gets genuine or not genuine

Copied codes caught when they turn up far away

Every check and alert on a live staff map

Demoed to the client's executive team

4live products: fintech, agritech, health, ops
107automated tests on the LendGrid prototype
0.911AUC on the CardioRisk prediction model
UTC+2overlaps Europe and US East mornings
Projects

Things I've built

LendGrid

Fintech · Lending platform

A lending operating system for Zimbabwean microfinance institutions. Loan officers assess applicants with a credit scorecard, managers review applications, and every decision lands in an audit log. A back-testing tool takes a lender's real loan book and reports how well the scorecard would have predicted defaults. The demo runs on a fictional lender's book and can connect to the full platform API.

  • ScoringCredit scorecard checked on screen and on the server
  • ValidationBack-testing with Gini, AUC and KS on a loan book
  • Quality107 unit and property-based tests
  • Next.js 15
  • TypeScript
  • Tailwind
  • Turborepo
  • PostgreSQL
  • Vercel

SeedShield

Agritech · Anti-counterfeit

Fake seed costs farmers whole harvests. Every bag carries a QR code: a farmer scans it and instantly sees whether the seed is genuine. If the same code is checked far from where it was first scanned, the system flags it as a copy. Company staff see every check and alert on a live map. Built and demoed for a seed company's executive team.

  • FarmersPhone check, points and tips, nearby dealers
  • DetectionCopied codes caught by location
  • StaffLive map dashboard and printable labels
  • Python
  • FastAPI
  • SQLite
  • Playwright
  • Railway

CardioRisk

Health · Machine learning

A web app that estimates a patient's heart-disease risk from clinical measurements and shows which factors drove the result, so the prediction is explained rather than a black box.

  • ModelAUC-ROC 0.911
  • ExplainedSHAP chart for every prediction
  • Python
  • scikit-learn
  • SHAP
  • Flask

Dissertation Tracker

Business ops · Web app

The app a small academic-services studio runs on: chapter deadlines with alerts, invoices and WhatsApp payment reminders, profit per project and a lead pipeline. Works offline, with end-to-end encrypted sync between laptop and phone.

  • Offline-firstData stays on the device
  • SyncEncrypted before it leaves
  • React
  • IndexedDB
  • Supabase
  • Vite

AI Contract Generator

Document automation · AI

A small business keeps its clients in a Google Sheet. Mark a row Ready and it becomes a finished, branded contract. An AI model rewrites the owner's rough notes into a professional scope of work, using only the facts given. Rows with missing details are flagged instead of producing a broken document.

  • Before20–30 minutes per contract, typed by hand
  • AfterA whole batch in under a minute
  • OutputGoogle Doc, Word and PDF for every client
  • Google Apps Script
  • Gemini API
  • Google Docs
  • Drive

Job Hunt Autopilot

Data pipeline · Gmail automation

Built for my own job search. Every morning it pulls remote jobs from four job boards, filters out roles I can't take from Zimbabwe, scores the rest against my skills and writes a tailored application for each. It reads my inbox every 15 minutes and only notifies me when an employer replies about an interview, while rejections are logged quietly.

  • Scans350+ new remote jobs a day
  • Keepsthe 5–10% that actually fit
  • Alertsonly on interviews and offers
  • Google Apps Script
  • REST APIs & RSS
  • Gmail
  • Google Sheets
  • Claude Code
What I build

If you can describe it, I can probably build it this week

Internal tools & small apps

Dashboards, trackers, admin panels and simple web apps your team actually opens.

e.g. order tracker, client portal, booking tool

AI-powered workflows

AI that reads, sorts, summarises and drafts, wired into the tools you already use.

e.g. inbox triage, document drafting, lead scoring

Automations & integrations

Connect Google Workspace, forms, email and third-party apps so data moves on its own.

e.g. Forms → Sheets → email, CRM sync, reports

Prototypes & MVPs

A working first version of your idea, fast enough to test with real users this month.

e.g. landing page + waitlist, AI chatbot, mini SaaS
Toolkit

AI-native, tool-agnostic

I build with whichever AI tools fit the job, and I pick up new ones quickly.

AI codingClaude Code · Codex · Cursor · GitHub Copilot · Lovable · Replit
AI modelsClaude · ChatGPT · Gemini
AutomationGoogle Apps Script · Zapier · Make · n8n
Built withNext.js · React · TypeScript · Python · FastAPI · Flask
Data & MLPostgreSQL · Supabase · SQLite · scikit-learn · SHAP
PlatformsVercel · Railway · GitHub · Google Workspace · REST APIs
How I work

From a plain-English problem to a working tool

  1. You describe the problemA message or a 2-minute screen recording is enough.
  2. I scope itWithin 24 hours you get what I'll build, the price and the date.
  3. I build with AIFast iterations, with a working version early so you can react.
  4. You get it runningSet up in your accounts, with a short video so your team can use it.
Work with me

Two ways to work together

Project

One tool, workflow or prototype at a fixed price agreed before I start. Small builds start around $100.

Part-time role

Join your team as the person who builds and maintains internal tools, AI workflows and automations.

Contact

Tell me what you'd like built.