Financial systems. Security tooling.

Financial systems & automation

My professional background is in fintech software engineering, test automation, and backend financial systems. That work spans APIs, FIX-related systems, CI/CD, and software delivery. My technical experience includes Python, Java, and SQL.

Trading workflows & test evidence

TradeFlow makes my approach to quality engineering inspectable: each test owns its authenticated session and order records, independent contracts validate responses, and cross-layer checks follow the exact record through a workflow. CreditFlow QE extends that evidence to a synthetic Java banking ledger with PostgreSQL transaction checks, independent API automation, and recorded browser/Jenkins execution. Both are synthetic engineering projects, not employer infrastructure or live financial services.

Evidence before automation

I keep observations, automated suggestions, and human decisions separate. In AegisGraph, deterministic detection and correlation establish case evidence. Replay exposes why an incident forms or which correlation conditions remain unmet. Rule comparisons show a proposed change across eight synthetic scenarios. Server-side checks validate AI claims and citations. Its hosted demo is read-only and uses a deterministic provider; live OpenAI analysis was validated in the local workflow. The fintech data is synthetic. The case study and article explain these boundaries, their tests, and the remaining production work.

I make failure explicit: a single invalid claim rejects the whole AI answer. An accepted answer still cannot change case state; that remains an analyst decision in the local workflow.

CreditOps AI explores the same need for review in trading-incident investigation. Its public showcase plays recorded model-assisted assessments using synthetic evidence, with no fresh model or MCP diagnostic calls during playback. Original proposals and recorded advisories remain inspectable; initiating causes remain unconfirmed. Its source is public, and the synthetic showcase runs on Railway.

The supporting security work applies the same discipline at other points in the lifecycle: validate scanner evidence before a CI decision, test detection changes against known cases, and require a matching rescan or reassessment before closing a finding. Those projects make their synthetic inputs and operational limits explicit.