I build research systems that make financial reasoning inspectable—from the source data and assumptions to the calculation, model and decision. My work sits at the intersection of financial analysis, quantitative research and production software engineering.
I am a computer science student at Arizona State University, currently focused on accounting, valuation, portfolio risk, credit, fraud and time-aware machine learning.
Working principle: a useful model should expose what it assumes, show where its evidence came from and remain honest when a simple baseline wins.
| Project | Research question | What it demonstrates | |
|---|---|---|---|
| 01 | QuantEdge Flagship · private |
How do filing evidence, market risk and valuation fit into one research workflow? | Multi-perspective equity research, scenario analysis, provenance and trust-labeled outputs. |
| 02 | LedgerLens | Do earnings, cash generation and balance-sheet changes tell a consistent story? | SEC filing normalization, three-statement relationships, capital efficiency and evidence exports. |
| 03 | IntrinsicLab | What must be true about a company’s economics for its valuation to make sense? | FCFF, WACC, terminal value, comparables and assumption sensitivity. |
| 04 | PortfolioPilot | Where does portfolio risk actually come from? | Risk contribution, downside analysis, optimization, costs and historical strategy evaluation. |
- CreditLens connects default probability, loss severity, exposure and loan pricing to calibration and threshold economics.
- FraudPulse evaluates fraud detection through missed-loss exposure, investigation cost and strictly prior account behavior.
- ChronosResearch tests whether market predictors survive purged, chronological out-of-sample evaluation against simple baselines.
QUESTION → ASSUMPTIONS → DATA CONTRACT → MODEL → VALIDATION → INTERPRETATION
visible traceable tested honest bounded
| Principle | In practice |
|---|---|
| Evidence before output | Preserve source tags, periods, retrieval context and versioned artifacts. |
| Uncertainty stays visible | Separate observed data, user assumptions, synthetic demonstrations and model estimates. |
| Time has direction | Fit preprocessing on the past, purge overlapping labels and reserve untouched future observations. |
| Baselines earn respect | Compare complex models with transparent alternatives and report when complexity does not win. |
| Calculations are contracts | Use typed boundaries, numerical checks and reproducible exports from engine to interface. |
| Domain | Tools I use |
|---|---|
| Financial research | Financial statements, DCF, WACC, capital efficiency, portfolio theory, credit and scenario analysis |
| Quantitative & ML | Python, pandas, NumPy, scikit-learn, TensorFlow, time-series validation, model evaluation |
| Applications & data | TypeScript, React, Next.js, FastAPI, Node.js, PostgreSQL, Prisma, REST APIs |
| Systems & delivery | Git, Docker, Linux, AWS, CI, test automation, data provenance and reproducible workflows |
- ReliScore — predictive-maintenance research across telemetry, temporal labels, training, inference and fleet triage.
- QuizBee — authenticated assessment workflows, grading and transactional persistence.
- Gridesign — responsive interface engineering and resilient service integrations.
Interested in financial research, risk systems and evidence-driven software.
Explore the repositories above or start a conversation.