Talks and sessions
Talks I can give at conferences, meetups and universities, each built on my public work, and the sessions I have led.
Talks I can give
Remote, or in person in Addis Ababa. Each one can run as a longer hands-on workshop.
Change data capture that tells the truth: lag, drift and deletes
Most CDC demos stop when the first row arrives. This talk starts there: how to tell a quiet table from a stopped connector, why deletes and ordering break naive pipelines, and how reconciling counts between the source and the warehouse catches the gaps a healthy dashboard hides. Built on two public pipelines, with the code on screen.
- For:
- Data engineers and platform teams running or planning change data capture.
- Length:
- Thirty to forty-five minutes, or as a workshop
Built from: World Bank CDC analytics, Kiva loan CDC analytics, CDC build guide, cdc-starter template
One CDC pattern, two builds: what changes between Airflow and Dagster
The same pipeline shape, from PostgreSQL through Debezium and ClickHouse to dbt, built twice with different orchestration and monitoring. What stayed the same, what each tool made easier, and which choices mattered more than the tool.
- For:
- Meetups, and teams choosing an orchestrator or a monitoring approach.
- Length:
- Thirty minutes
Built from: The pattern, stage by stage, World Bank CDC analytics, Kiva loan CDC analytics
Is your data ready for AI? The checks before the model
The questions I would settle before funding an AI use case: who owns each source it needs, whether the definitions agree, what personal data the model will see, what it will cost to run, and how its quality will be watched once people rely on it.
- For:
- Leaders and teams planning AI on top of the data they already have.
- Length:
- Thirty minutes
Built from: AI readiness checklist, AI readiness check
Choose the shape before the tools: warehouse, lake or lakehouse
A few choices set most of a data platform's cost and risk before any code is written: the storage shape, the integration style, who owns the data, and where definitions live. How I make each one, and the signals that should change the answer.
- For:
- Architects, engineering leads, and university audiences.
- Length:
- Thirty to forty-five minutes
Built from: The choices I make first, Platform choice scorecard
Sessions I have led
Google Developer Student Clubs Lead
Wolaita Sodo University, Aug 2022 to Jul 2023
Led the university chapter of Google Developer Student Clubs and hosted hands-on events and workshops on machine learning, web and mobile development. Google recognized the year with a letter of commendation.
Internal machine learning lectures
Kifiya Financial Technology (internal), Apr 2026
Gave internal lectures on gradient-boosted trees (XGBoost) to colleagues on the data teams.
Invite me to speak
Tell me about the event, the audience and the date, and I will reply by email.
Write to me