I build practical software that turns data + automation + AI-assisted analysis into reliable, inspectable workflows.
| Project | What it demonstrates |
|---|---|
| 🔎 Evidence-driven agent skills | Reusable AI-agent workflows for auditing, certification, debugging, evidence reporting, tenant-isolation testing, verified backups, safe delegation, and non-destructive changes. |
| 📈 Market Intelligence Platform | Local-first research platform with Python, SQLite, scoring, news normalization, mock AI analysis, alerts, automation, health checks, and a read-only dashboard. |
Evidence-driven agent skills
9 reusable skills · adversarial reviewer · synthetic examples · automated validation · CodeQL · MIT
Market Intelligence Platform
Python · SQLite · immutable research history · provider abstractions · CI · security policy · human-reviewed alerts
| Area | Focus |
|---|---|
| Automation | Python workflows, schedulers, alerts, integrations |
| Data | SQLite, validation, append-only history, structured pipelines |
| AI systems | Agent workflows, research assistance, evidence-based review |
| Interfaces | Local dashboards, viewers, JSON APIs |
| Reliability | Tests, smoke checks, CI, CodeQL, secret protection |
| Research tooling | Market monitoring, scoring, news intelligence, portfolio research |
Automate repetitive work, not accountability.
I prefer deterministic pipelines where deterministic logic is enough, keep important decisions inspectable and human-controlled, and separate observations, derived metrics, assumptions, and AI-generated analysis.
- Building automation and integration systems that are useful in real workflows.
- Expanding AI-agent tooling around review, research, validation, and delegation.
- Developing data and market-research tools with clear boundaries between evidence, scoring, alerts, and human decisions.
- Publishing sanitized, reproducible open-source versions of larger private projects without exposing operational data or credentials.