Dar es Salaam · 2019 to 2022
Biomedical equipment
How medical devices work, and what it takes to trust what they measure.
Data Scientist & Software Engineer
Open to 2027 internships & roles
From Dar es Salaam to data science
I trained as a biomedical equipment engineer. Now I clean up data that doesn't agree with itself.
I'm a data science student at BYU-Idaho. At work I evaluate transfer credit: 300+ academic records a semester from schools that each describe the same course differently. Most of what I know about messy data, I learned there.
- Degree
- BS Data Science, CS minor, BYU-Idaho · expected July 2027
- GPA
- 3.90
- Records evaluated
- 300+ per semester, 100% accuracy
- Based in
- Rexburg, Idaho · originally from Dar es Salaam, Tanzania
01 · About
Most of my work starts
Most of my work starts
where two records disagree.
- Now
- BS Data Science (CS minor)
BYU-Idaho
Expected July 2027 - Before
- Diploma, Biomedical Equipment Engineering
Dar es Salaam Institute of Technology
2019 to 2022
I studied biomedical equipment engineering in Dar es Salaam. The hard part was rarely one device. It was everything around it that had to be right before you could trust a reading. I came to the U.S. for computer science, switched to data science, and kept running into the same kind of problem.
My campus job is transfer credit evaluation. Every semester, hundreds of transcripts come in from schools that name, number and weight their courses differently, and someone has to decide what counts as what. It's basically schema matching, except a student's graduation date depends on getting it right.
So I build tools for the tedious parts: pulling data out, matching it, and checking it. At a hackathon I built a tool that reads web pages so researchers don't have to open every one. At work I reordered our verification steps and cut processing time by 20%. It's not the exciting part of data work, but everything after it depends on it.
How I got here
Different fields,
Different fields,
same kind of problem.
In each one, the job was making sure the pieces agreed before anyone relied on the result.
Moving to the U.S. · 2022 to 2024
Software
Learning to turn manual, step-by-step work into code people can reuse.
Rexburg · Now
Data
Matching and cleaning up records before anyone analyzes them or makes decisions with them.
Experience
Where I've worked
Apr 2025 to Present
Transfer Evaluation Processor
Student Records & Registration · BYU-Idaho
Evaluate 300+ academic records per semester with 100% accuracy, matching courses between schools whose catalogs don't line up. Reordered the verification steps, which cut processing time by 20%. Resolve 50+ data discrepancies a term by tracing each one back to the department that created it.
Sep 2023 to Apr 2025
Testing Center Proctor
Student Testing Center · BYU-Idaho
Ran exam security procedures for 50+ test-takers at a time and helped 20+ people a week with technical problems.
Toolkit
Tools I use
- Data & ML
- PythonpandasNumPyscikit-learnTensorFlowHugging FaceR
- Data systems
- SQLPostgreSQLData wranglingETLDockerLinuxGit
- Engineering
- JavaScriptC#AstroChrome Extensions
- Coursework
- Data StructuresAlgorithmsStatisticsDatabase ManagementCalculus
02 · Selected work
A few things I've built
Interactive demo · made-up data
How mismatched course records get matched
This demo uses made-up records to show the kind of mismatch I deal with at work: the same course listed under different titles, credit values or department codes.
Show mismatch
Incoming catalogsinconsistent
Our catalogverified
No records in this view.
001
Mar 2025 · Hackathon, UVU
SumAI — AI Extraction Pipeline
- Problem
- Researchers waste a lot of time opening pages just to figure out whether a source is worth reading.
- Approach
- A Chrome extension with a Python backend that extracts and processes the active page, then lets researchers ask page-specific questions without leaving the tab.
- Outcome
- Cut research time by 80% in our testing. Built start to finish in 24 hours at a Utah Valley University hackathon.
PythonpandasSeleniumChrome Extensions APILLM summarization
002
2025 · BYU-Idaho
Transfer Evaluation Workflow
- Problem
- 300+ transfer records a semester, each from a school with its own catalog, course names and credit system. Matching them by hand was slow, and a mistake can push back a student's graduation.
- Approach
- Moved the checks that catch the most errors to the start of the process, and set up one standard way to flag problems and send them to the department that can fix them.
- Outcome
- Processing is 20% faster with accuracy still at 100%, and 50+ discrepancies a term get resolved instead of escalated.
SQLData verificationProcess designCross-team coordination
003
2026 · Open-source project
FamTree — Interactive Family Knowledge Graph
- Problem
- Family histories become difficult to navigate and update as generations and relationships grow, while the underlying information still needs strict privacy controls.
- Approach
- Built a responsive graph interface with search, timelines and member profiles, backed by invite-only Firebase access and an administrator review queue for proposed changes.
- Outcome
- Turned a static family-history document into a structured, explorable system with reviewable edits. The source is open for contributions; all public examples use invented people and no real family records.
ReactTypeScriptD3FirebaseZustandVitest
03 · Contact
If your data doesn't line up,
If your data doesn't line up,
I'd like to hear about it.
I'm looking for data science and data engineering internships, and full-time roles starting in 2027.
I enjoy the early part of the pipeline most: getting data out, matching it, and making sure it's right. Either way, feel free to reach out. I reply to every message.