Peter Tinashe Mundowa
Data Scientist and AI Engineer | Harare, Zimbabwe
petermundowa25@gmail.com | petermundowa.com | LinkedInlinkedin.com/in/petertinashemundowa | GitHubgithub.com/PeterTheDataScientist | Zindizindi.world/users/PeterTheAnalyst
Profile
I turn undocumented enterprise data into measurable business value. As a Data Scientist and AI Engineer at AI Creative Solutions, I build ERP data pipelines, lakes and warehouses, production platforms, machine learning models and governed AI agents. My independent research asks why models that excel in well-resourced conditions fail in real ones; it won Best Poster at Deep Learning IndabaX Zimbabwe 2026, with selection to represent Zimbabwe at the Deep Learning Indaba 2027. Featured in ITU News (June 2026) and Zindi Ambassador for Zimbabwe.
Experience
Data Scientist and AI Engineer, AI Creative Solutions, HarareJan 2026 to present
- ERP and accounting data engineering on SAP Business One, Sage and a PostgreSQL-based ERP: incremental loads from read-only standbys into lakes and warehouses modelled as facts and dimensions, with schema-drift checks, quarantine, and reconciliation that must tie before any cut-over.
- Production platforms (React, TypeScript, Express, PostgreSQL, Drizzle ORM) that replace legacy databases and spreadsheets: migrations where every source row is accounted for, role-based access, envelope encryption of personal data and append-only audit trails, guarded by automated test suites that include security regression tests against real PostgreSQL.
- Governed AI agents on live business data: permission-gated tool registries, read-only SQL that the server constrains and scopes, human approval before consequential actions, and an audit record for every AI call.
- Analytics on top of ERP data: management packs with reconciliation checks that must tie, debtor and creditor ageing, RFM segmentation and churn risk, market-basket analysis, stock cover and cash forecasts, served on screen and as spreadsheets from one model.
- Sectors: manufacturing, agriculture, healthcare and distribution, in Africa and beyond.
Revenue Assurance Intern, Steward Bank, HarareAug 2024 to Aug 2025
- Revenue-assurance dashboards and automated data extraction across billing systems.
- Automated the Reserve Bank of Zimbabwe exchange-rate capture, cutting a four-hour manual task to 15 minutes.
Selected projects
- BroilerVision: computer vision for broiler houses (live). An overhead camera, a detector and a dashboard for the stockperson: it counts broilers with an honest range and flags any bird that stops moving. On held-out data it finds 82% of birds, 84% of its boxes are real, and it runs at 83 ms a frame on CPU. broiler-vision.streamlit.app
- Internal knowledge assistant: answers with sources, or an honest refusal (live demo). Answers staff questions only from company documents and cites every claim: 89.6% of 48 labelled questions passed, 100% citation precision, zero safety failures (Llama 3.1 8B). github.com/PeterTheDataScientist/internal-knowledge-assistant | Live demo: grounded-knowledge-assistant.streamlit.app
- Tomato leaf disease detection: hybrid CNN-Transformer (live demo). My completed BSc Honours dissertation, deployed as a live demo: a hybrid CNN-Transformer for budget phone photos that raised low-light F1 from 0.42 to 0.78, a 36-point gain. github.com/PeterTheDataScientist/Tomato-Leaf-Disease-Detector | Live demo: tomato-leaf-disease-detector-peter.streamlit.app
- ZimCropGuard: multi-crop disease detection under domain shift (ongoing). Research on crop disease models that hold up under real African field conditions: EfficientNetB0 and ViT-B/16, fused, across five staple crops, from one phone photo. petermundowa.com/work/zimcropguard
- Grid Atlas: electricity availability from night-time satellite imagery (preview). Night-time electricity availability for 26 African cities from VIIRS imagery, each city measured only against its own history, in a public dashboard. peterthedatascientist.github.io/grid-atlas
- Mapping equity: the flood crossings an open map cannot see (completed). My Bias Bounty writeup: 8,454 official Texas low-water crossings, none marked as flooding on the open map, and 216 areas cut off, with an exact, reproducible build of the coverage gap score. github.com/PeterTheDataScientist/bias-bounty-coverage-gap
- Also in preview: Zimbabwe Crop Atlas (crop monitoring through the cloud season, github.com/PeterTheDataScientist/zimbabwe-crop-atlas) and BorderFlow (corridor state for Southern Africa, github.com/PeterTheDataScientist/borderflow).
Awards and recognition
- Best Poster, Deep Learning IndabaX Zimbabwe 2026 (ZimCropGuard), with selection to represent Zimbabwe at the Deep Learning Indaba 2027
- Featured in ITU News, Jun 2026: "AI for Good: Meet the people powering machine-learning solutions"
- Quoted in University World News, Jul and Aug 2026, on education financing in Zimbabwe, for the Right for Education Zimbabwe Chapter
Competitions
Zindi: 2 gold, 6 silver and 7 bronze medals across 41 challenges; best global rank 99 on a platform of 100,000+ AI builders (November 2026). Selected final placings:
- 3rd of 85 ITU Measuring What Matters Proposal Challenge (AI environmental footprint), Oct to Nov 2025
- 3rd of 10 Face Spoofing Detection Hackathon, IndabaX Zimbabwe, Nov to Dec 2025
- 8th of 47 Maize Crop Disease Detection Challenge (team), Jul to Aug 2025
- 10th of 40 IndabaX Zimbabwe 2026 Hackathon: Loan Default Prediction (team), May 2026
- 35th of 364 DigiCow Farmer Training Adoption Challenge, Jan to Mar 2026
- 52nd of 288 The African Trust and Safety LLM Challenge, Mar to Apr 2026
Leadership and community
Zindi Ambassador for ZimbabweAug 2025 to present
Organising events, running competitions and mentoring practitioners. Mentor to final-year data science students on their capstone projects.
Right for Education, Zimbabwe ChapterFeb 2024 to present
Author (Feb 2024 to present), Vice President (Aug 2024 to Oct 2025), President (Oct 2025 to May 2026).
Speaking
- Speaker, R:Ed Africa Leadership Accelerator, Q3 series, 11 and 12 Sep 2026: "Leading Without a Title: Building Authority Through Work, Not Position"
- Speaker, Deep Learning IndabaX Zimbabwe 2025: maize leaf disease detection, in the demo and startup session
- Recorded talk, 11 Dec 2025: "Leadership in the Age of Technology: Guiding Change with Wisdom", presenting my R:Ed article of the same name
Education
BSc Honours Degree in Data Science and Informatics, Upper Second Class (2.1), University of Zimbabwe2026
Dissertation: a hybrid CNN-Transformer for tomato leaf disease detection, robust to low light and budget phone cameras.
Certifications
- Microsoft (2025, renewed 2026): Azure AI Engineer Associate; Azure Data Scientist Associate; Power BI Data Analyst Associate
- DataCamp certifications (2024 to 2025): AI Engineer for Data Scientists, Data Engineer, Data Analyst, SQL and Data Scientist, all at Associate level
- Anthropic (2026): seven courses, including Building with the Claude API, Claude Code in Action and Introduction to Model Context Protocol
- Programmes: McKinsey Forward Program (Dec 2024); Aspire Leaders Program (Jan to Mar 2025); WorldQuant University Applied AI Lab (Nov 2024 to May 2025); WorldQuant University Applied Data Science Lab (Nov 2023 to Apr 2024); YALI Fundamentals of Responsible Leadership (Aug 2024)
Skills
- Languages and data: Python, SQL (T-SQL, PostgreSQL), pandas, NumPy, Apache Spark, DuckDB
- Data engineering: ETL and ELT pipelines, incremental loads and change capture, data quality and reconciliation, dimensional modelling, data lakes and warehouses, Apache Airflow, dbt, Snowflake, Databricks
- Business systems: SAP Business One, Sage, Microsoft SQL Server
- Machine learning: PyTorch, TensorFlow and Keras, scikit-learn, LightGBM, XGBoost, Hugging Face, YOLO, OpenCV, SHAP, evaluation under domain shift
- AI engineering: governed AI agents, safe natural-language-to-SQL, RAG with FAISS, Model Context Protocol, the Claude API, MindStudio, FastAPI, Docker, AI governance (EU AI Act, NIST AI RMF, ISO/IEC 42001)
- Software: TypeScript, React, Node.js, Express, Drizzle ORM, Astro, Streamlit, pytest, GitHub Actions, role-based access, encryption and audit trails
- Analytics: Power BI with DAX, forecasting, RFM segmentation, market-basket analysis, anomaly detection
- Cloud and geospatial: AWS, Microsoft Azure, Vercel, Sentinel-1 and Sentinel-2, Landsat 8 and 9, VIIRS, Digital Earth Africa, rasterio, geopandas, GDAL