Open to Collaborate
"If we can detect this early, that means increased lifetime for this patient."
Interview and article by Pamela Pont
Kelvin Mutua is a PhD student in Kenya building cancer risk and survival models, in a country where the local data to train them barely exists. He is looking for collaborators who can help close that gap.
Open to collaboration. Kelvin is a data science and risk modeling PhD student focused on predicting health outcomes. He modeled cancer patient survival for his master's, and his doctoral work builds AI-driven early warning systems for population health risk. He is on the Research Exchange to explore the open gynecologic cancer data sets.
Where he is in the program: Now in his second year, with coursework complete and the research phase running through year three. He works within a funded multi-institution program with supervisors and principal investigators in place, so he is not looking for a program or an advisor. The program includes the Technical University of Denmark, the University of Nairobi, Chuka University, the Kenya Bureau of Standards, and the African Centre for Technology Studies, and supports six PhD students and five master's students across data science, microbiology, agronomy, fermentation, and social research.
What Kelvin is looking for
Collaborators on cancer risk and survival modeling. Kelvin wants to work with researchers on gynecologic cancer outcomes who could use a modeling partner, and on projects where his methods can be applied to existing data sets, including co-authored papers and joint work. Entering his research years, he has the time to commit and every incentive to see a collaboration through to publication.
Data, in both directions. Kenya has very few local cancer data collection points, so most modeling there depends on data from abroad. Kelvin is using the Exchange to reach open-access data sets he could not otherwise get to. He also wants to tackle the harder problem: what it would take to bring Kenyan cancer data onto a shared platform in a way that respects data sovereignty and local hosting requirements.
Mentorship with reach. A mentor who is well connected in the field and can help him build a network alongside the research itself.
Pathways to real-world use. Kelvin wants partners who can help move findings into policy, clinical and public health practice, and tools that actually get used.
Peer connection. He would welcome introductions to other PhD researchers on the Exchange and opportunities to present the work at conferences.
What Kelvin brings
Survival and risk modeling experience. His doctoral work combines simulation and imputation methods, bootstrapping, machine learning, distributional modeling, and meta-learners that combine models to improve accuracy, all trained on local data rather than approaches imported from elsewhere.
A working early warning system, end to end. His dissertation work is producing a full early detection pipeline for aflatoxin contamination in Kenya: predictive models, a national geo-referenced database, and an app that generates real-time signals. The architecture is deliberately built to generalize.
Applied modeling from outside academia. Before his PhD, Kelvin worked in market research and consulting, including exposure assessment for clients. It taught him that methods move faster in the field than in the classroom.
Ground-level knowledge of the Kenyan data landscape. He knows where the collection gaps are, which institutions hold what, and how hard the access conversations get. He has raised the under-recognition of women's cancers publicly, including at a University of Michigan seminar hosted at Kenyatta National Hospital.
Background
Kelvin's interest in cancer began with his master's thesis on patient survival. The obstacle he met there has shaped everything since: models built on data from elsewhere cannot be fully trusted for the people they are meant to serve. He sees the fix as infrastructure, not just analysis.
So he built that infrastructure first in an adjacent domain. Aflatoxin contamination is a persistent threat to Kenya's food supply and public health, and the system he is developing for it is a template as much as a product. His goal is to carry it into cancer, where predicting risk earlier changes what treatment can accomplish.
In his own words
What are you currently focused on in your PhD?
My doctoral research focuses on data science and risk modeling, with an emphasis on health-related data including cancer, malaria, and aflatoxin contamination. Building on this foundation, I aim to integrate AI to develop early warning systems: first for aflatoxin contamination in Kenya, and then with the potential for early cancer detection, which extends the impact considerably.
What kind of support or collaboration are you hoping to find in the Research Exchange?
Resources are the first thing, alongside people who can inform the work as it develops. Beyond that, I want what I build to reach people who can act on it, because too much research ends up in a journal or as a printed document on a shelf. I would also value connecting with other PhD researchers working in this space, and opportunities to present the work at conferences.
What questions or problems in your field do you find yourself most curious about right now?
Two challenges in particular. The first is data accessibility. Kenya does not have enough local cancer data collection points, so most modeling relies on data from abroad. Building a national, geo-referenced database is one of my first objectives, and the access conversations with institutions have been the hardest part. The second is computational power. Certain models are computationally intensive and can take a significant amount of time to run, which limits how much I can iterate. Both are areas where collaboration would be especially valuable.
When you imagine your research a few years down the line, what does that look like?
I envision a fully developed early warning system: the models themselves as well as a complementary app that generates real-time signals and confirms the system is functioning reliably. Alongside this, I see myself contributing to a national, federated database, an open-data platform where verified organizations can securely upload and access data, with processes in place to maintain integrity and trust.
Beyond the food safety application, I envision extending this framework to early cancer detection, with the goal of predicting the presence of cancer before a patient presents at a hospital. If we can detect this early, that means increased lifetime for the patient.
What kind of collaborator or mentor would make the biggest difference for you?
Someone highly responsive and genuinely invested in my growth, well established in the field, and able to help open doors through networking. I see mentorship as taking two possible forms: helping me reach a destination I have already envisioned for myself, or collaborating directly on research. Above all, I value honesty. If I am taking on more than I can realistically handle, I want a mentor who will tell me so directly and help me recalibrate, while staying encouraging throughout.
Is there anything outside of academics that has shaped your approach to your work?
Before pursuing my PhD, I worked in market research, and that shaped my approach significantly.
For example, I was working on a simulation model that required imputing missing consumption measurements. I initially used a normal distribution and realized it was not producing reliable results. That sent me back to the literature to see what methods others were using, and my supervisor pointed me toward alternatives beyond simple techniques like dividing the limit of detection by two: machine learning algorithms, linear regression, and other distributional methods.
That experience instilled the habit of looking at what others have done and asking how those approaches might be incorporated into my own research.
What surprised you about pursuing a PhD once you actually started?
How non-linear the process really is. There have been moments when I felt confident I had completed a piece of work, only for my supervisor to point out that the underlying research was not yet thorough enough, which sent me back to strengthen the foundation before moving forward. I have also been struck by how much literature there is to engage with.
Get in touch
He is seeking collaborators, mentors, and data partners for his research years. If you work on gynecologic cancer outcomes, survival or risk modeling, or cancer data infrastructure, or if you can help move findings toward policy and practice, he would welcome the conversation.