AI · DATA SCIENCE · QUANTITATIVE FINANCE
I’m Rancy Chepchirchir — an AI researcher, data scientist and quantitative finance practitioner working across trustworthy AI, financial machine learning, scientific machine learning, computer vision and NLP.
Trustworthy AI, foundation models, uncertainty, financial event sequences & model governance.
THE THREAD
Different domains. Same question: how do we build intelligent systems that remain useful, interpretable and reliable when uncertainty is part of the problem?
SELECTED RESEARCH
Research spanning quantitative finance, trustworthy machine learning and AI systems that must operate under uncertainty.
A comparative investigation of deepfake detection architectures, examining how different deep-learning approaches perform on manipulated-media classification and what their trade-offs imply for robust, trustworthy detection systems.
Brain-tumour segmentation for Sub-Saharan African MRI data, combining nnU-Net, cross-domain evaluation and generative style-transfer augmentation to improve robustness under lower-resource imaging conditions.
Studying covariance estimation under market microstructure noise and evaluating estimators from the portfolio-selection perspective, including QMLE and high-frequency return constructions.
Comparing PINNs, DeepONets and classical finite-difference methods for derivative pricing, with particular interest in American options, PDE constraints and uncertainty-aware approximation.
Empirical analysis of futures introduction and spot volatility across agricultural commodities and markets using ARCH, GARCH and EGARCH modelling.
FEATURED WORK
A curated selection of systems, experiments and research prototypes rather than an exhaustive repository list.
Selected cards expose methods, measured results and model behaviour directly — not just project descriptions.
Comparative deep-learning research on manipulated-media detection, asking not only whether a detector performs well, but what architecture choices mean for robust deployment.
Glioma segmentation under domain shift, combining nnU-Net experiments with generative style-transfer augmentation for Sub-Saharan African MRI data.
QUANTLAB / RESEARCH WORKSTATION
An interactive research environment connecting classical derivatives pricing with scientific machine learning. QuantLab moves from Black–Scholes and CRR through finite-difference solvers, Monte Carlo simulation and volatility calibration to American-option PINNs, error topography and free-boundary analysis.
Applied NLP and visual-language work including Kenyan Sign Language and Swahili.
View project ↗Representation learning for financial event streams and uncertainty-aware decisions.
Discuss research ↘A full-stack knowledge system for books, papers, notes, writing, semantic retrieval, research discovery and connected ideas — powered by Next.js, PostgreSQL/pgvector and AI-assisted search.
ABOUT
My path into AI began in financial economics and mathematical finance, where uncertainty is not a nuisance to remove but a quantity to model.
That perspective still shapes how I approach machine learning: I care about what a model predicts, but also why it predicts it, how reliable that prediction is, how it behaves when the world changes, and what happens when people rely on it.
My work has moved through quantitative finance, data science, NLP, computer vision, medical imaging and AI safety. Across those areas, the underlying question remains surprisingly consistent: how do we build intelligent systems worthy of the decisions we ask them to make?
Reader of philosophy and literature. Student of Stoicism by inclination. Builder of personal knowledge systems by habit.
HOW MY RESEARCH CONNECTS
Hover or focus on a node to see the projects and ideas it connects to. The common thread is reliability under uncertainty.
EXPERIENCE & EDUCATION
Jan 2024 — Present
LinkedIn Learning
Technical education and applied data-science content, including Python for finance.
Apr 2023 — Feb 2025
Derivation LLC
NLP systems, data applications, model development and applied machine-learning workflows.
May 2024 — Jul 2025
University of Hull
Deepfake detection, deep learning, computer vision and trustworthy machine learning.
2021 — 2022
Maitri Capital
Financial research, quantitative analysis and data-driven investment work.
Apr 2018 — Jun 2020
WorldQuant University
Derivatives, econometrics, risk, portfolio theory and computational finance.
Jan 2011 — Jun 2015
Strathmore University
Economics, finance, statistics and quantitative methods.
RESEARCH INTERESTS
SELECTED PUBLICATIONS
Selected papers spanning trustworthy AI, medical imaging and quantitative finance.
SELECTED RECOGNITION
A small selection of conferences, programmes and awards that mark important stages in my research journey.
Paper accepted with a poster presentation in medical imaging research.
Received LACUNA Fund support to attend MICCAI and presented a poster.
Won both Most Progress and Best Team.
Accepted into the medical-imaging training programme.
Paper accepted for presentation in Liverpool.
Mentor with Coding Black Females, supporting students and early-career technologists.
CURRENTLY
The portfolio is the archive. This is the moving part.
Robustness, provenance, uncertainty and governance for systems that operate across modalities.
Two evolving research systems: QuantLab for computational finance and scientific ML, and Codex for books, papers, notes, semantic retrieval and connected ideas.
Not merely whether a model is confident, but whether its confidence means what we think it means.
Machine learning, finance, research field notes, philosophy and whatever refuses to leave my head.
WRITING
Technical notes, project stories and occasional field notes — on machine learning, quantitative finance, research, people and the questions that stay with me.
A practical exploration of binomial trees, finite-difference methods and Monte Carlo approaches for valuing American options — an early thread in my continuing work on computational finance and scientific machine learning.
Read on Medium ↗Building a real-time sign-language detection pipeline with TensorFlow, SSD MobileNet V2 and OpenCV — from data collection and annotation through inference and practical limitations.
Read on Medium ↗Reflections from Deep Learning Indaba in Ghana: a week of African AI research, community, conversations and the kind of ideas that continue long after the conference ends.
Read on Medium ↗BEYOND THE MODELS
When I’m not training models or solving PDEs, you’ll find me reading, writing, playing the guitar, watching football, travelling, and chasing ideas that do not always fit neatly into a dataset.
Music keeps a different kind of time — less measurable, still worth learning.
Philosophy, literature, history and books that leave behind better questions.
Sometimes technical, sometimes personal, always an attempt to understand or remember.
LET’S TALK
I’m interested in conversations around trustworthy AI, AI safety, financial ML, scientific machine learning, research collaborations and ambitious technical projects.
Coffee is also an acceptable protocol.