FinTech · ML & AI Specialist · Accounting · Blockchain
FinTech, Machine Learning and AI Specialist with deep accounting expertise and blockchain knowledge. Building robust, production-grade systems across financial workflows, quantitative risk models, agentic pipelines, and decentralized protocols.
ABOUT
I am a FinTech, ML and AI specialist pursuing an M.S. in Financial Technology at the University of Central Florida, holding a B.Sc. (Hons) in Applied Accounting from Oxford Brookes University. My dual technical and financial grounding enables me to view financial architecture holistically—from balance sheet internal controls, ERP ledgers, and audit trails to deep learning models, agentic workflows, and distributed ledger mechanics.
My work focuses on architecting end-to-end data pipelines, developing machine learning solutions with strict guardrails, extracting actionable intelligence from regulatory filings via LLMs, and securing smart contract protocols.
Beyond algorithmic modeling, I bring real-world entrepreneurial experience as Co-Founder of Charles Beatina PBC. Managing budgets, supplier relations, cash flow, and business operations ensures my technical systems address practical financial realities, not just idealized benchmarks.
Python · PyTorch · Scikit-Learn · LLMs & Ollama · LightGBM · Graph Neural Networks · SQL · Solidity · Full-Stack Development
SELECTED WORK
An interactive quantitative fixed-income laboratory designed for bond valuation, yield curve modeling, interest rate sensitivity, and duration/convexity risk analysis.
End-to-end Python pipeline extracting corporate product releases and material events from SEC Form 8-K filings via EDGAR API and Ollama LLMs into structured CSV datasets.
An AI-powered tax workflow orchestration platform combining automated tax question answering with rule-based task routing, document handling, and deterministic financial guardrails.
Applied quantitative risk management platform for evaluating downside exposure, scenario stress-testing, and dynamic hedging simulations paired with automated reporting.
Probabilistic market analysis engine using LightGBM and engineered distribution features to classify equity return environments across 5 discrete regimes with explicit uncertainty estimation.
Applied ledger anomaly detection platform identifying accounting errors and fraudulent journal-entry patterns by fusing domain accounting rules with supervised and relational learning models.
Multimodal risk-prediction framework fusing structured financial metrics with narrative MD&A disclosures to assess corporate insolvency and forward distress probabilities.
Graph and MLP deep learning architecture leveraging Bitcoin on-chain transaction network chainlets as structural signals for dynamic price and volume forecasting.
Synthetic ERP accounting dataset benchmark designed for training fraud detection, reconciliation gap analysis, authorization bypass, and audit trail anomaly models.
CAPABILITIES
Accounting controls, ledger analytics, variance & reconciliation analysis, corporate budgeting, financial modeling, portfolio risk analysis, fixed-income modeling, regulatory filings (SEC EDGAR).
PyTorch, TensorFlow, Scikit-Learn, LightGBM, XGBoost, Graph Neural Networks (GNNs), LLM pipelines (Ollama), prompt chaining, deterministic agentic routing, guardrail design.
Solidity, Ethereum smart contract security & opcode analysis, on-chain transaction graph analytics, Python, SQL, R / RStudio, HTML5, JavaScript, modern full-stack web integration.
ENTREPRENEURSHIP
Co-founded and managed business and financial operations for a processing and distribution enterprise. Directed capital expenditure, monthly budgeting, vendor negotiations, procurement cycles, disbursements, and cash runway management.
This operational tenure guarantees that every model, algorithm, and software system I construct is anchored in practical commercial utility and cash-flow discipline.
RESEARCH
Comparative empirical research on structured, temporal, and relational learning architectures designed for automated general ledger validation and anomaly detection.
Comprehensive benchmarking framework combining opcode extraction, static analysis, and graph neural network representations across Ethereum smart contract vulnerability datasets.
ACADEMIC REFERENCES
Assistant Professor of Computer Science, FinTech and Digital Forensics.
mst.ozdag@gmail.com ↗Lecturer, Department of Banking and Finance / Actuarial Science, NUST.
lungisani.mpofu@nust.ac.zw ↗CONNECT
Available for quantitative FinTech roles, machine learning engineering, and institutional research collaborations.