AI Engineer | AI Quality / QA Engineer | AI SDET | Software Engineer
I build AI applications and the software that tests them. My work spans RAG, browser agents, LLM evaluation, test automation, APIs, and interactive web applications. I bring a software QA background, including functional, regression, and compliance testing at iTech Labs and Gaming Labs International, to projects where behavior needs to be measurable and failures need to be explainable.
- AI engineering: Document retrieval, source-linked answers, model integrations, and tool-using browser agents.
- AI quality and QA: LLM-judge calibration, deterministic output checks, adversarial test cases, and traceable evaluation reports.
- AI SDET and test automation: Playwright and Pytest frameworks, regression coverage, contract and performance testing, and CI.
- Software engineering: Local-first services, browser interfaces, validation, error handling, and automated checks.
| Project | What I built | Stack |
|---|---|---|
| Kit | Local RAG starter for TXT and Markdown documents, with passage chunking, hybrid semantic/keyword retrieval, optional OpenAI embeddings and cited answers, and persistence. Offline mode shows exact excerpts without AI generation. | Node.js, JavaScript, OpenAI APIs |
| ReleaseRadar | Regression-testing prototype with seeded checkout bugs, frozen assertions, exportable evidence, an AI-assisted QA copilot, and an API/testing workbench. | JavaScript, Playwright, OpenAI, Vercel, Supabase |
| Assay | LLM-judge calibration against human-labeled examples using Cohen's kappa. The default gate blocks uncalibrated scoring; reports retain calibration provenance and model-call usage. | Python, Ollama, LLM evaluation |
| LLMTestKit | Local-model evaluation and red-team tooling with deterministic checks, a correctness judge, prompt-injection and jailbreak cases, and HTML reports. | Python, Ollama, Pytest |
| Jev Flight Agent | Natural-language flight-search prototype: Jev selects browser operations, a separate model supplies field text, and Playwright executes the actions. Stops at search results without booking. | Python, Jev / TypeSafe AI, Playwright |
| Playwright Test Automation Framework | SauceDemo UI automation with page objects, JSON-driven tests, failure screenshots, parallel execution, and HTML reports. | Python, Playwright, Pytest |
Explore: Portfolio · ReleaseRadar demo · ReleaseRadar setup and verification notes
These are portfolio projects and prototypes. Kit's automated provider tests use mocks; citation IDs are validated, but factual support is not automatically verified. ReleaseRadar's live integrations have documented setup and verification limits. Assay and LLMTestKit are early-stage tools, with broader RAG and agent evaluation on their roadmaps. Jev Flight Agent requires local Chrome and API keys.
-
Orbit Desk Companion · Interactive demo
A Three.js desk-companion concept with a procedural droid and planet, scroll-controlled assembly, five device simulations, and a circuit workshop. Includes keyboard and reduced-motion support plus automated checks. Sensors, voice, and device behavior are simulated; the hardware design is not physically validated. -
ArambyaAIAgent · Public demo
A local discovery and day-planning prototype that retrieves public place, weather, and event data, streams tool activity, and exports a sourced itinerary. The public planner is rule-based, not LLM-powered. The local version includes an optional model-ranking adapter.
More testing, AI, and MLOps projects
| Project | Focus |
|---|---|
| Locust Performance Suite | Load testing a Flask application with p95 latency assertions and generated reports. |
| Pact Contract Testing | Consumer-driven API contract testing with Python, Flask, and Pytest. |
| Oath | OpenAPI dependency graphs, boundary tests, and violation ranking by downstream impact. |
| Onset | Repeated test runs and statistical comparisons to investigate when a test became flaky. |
| AI Quality Ops Platform | Quality-pillar test plans, policy thresholds, reports, and a bundled mock-service demo. |
| AI Agent Validation Framework | Mock-agent tests with required/forbidden phrase checks and TextBlob sentiment checks. |
| AI Agent Quality Sandbox | Doer, adversary, and judge scenarios with rule-based checks and simulated performance. |
| Medical Q&A with RAG | A medical-information RAG prototype using WHO source material, Chroma, FastAPI, and Hugging Face inference. |
| Synapse | LangGraph-based agent experiments with tools and memory. |
| Model Hub | Text and image moderation API experiments. |
| Fraud Detection | Fraud-detection API and drift-monitoring experiments. |
| OmniServe | Early local-first multimodal assistant work. |
I'm always open to discussing software quality, test automation, AI reliability, or potential collaborations. Reach out!
"AI that works is AI that is tested, monitored, and continuously improved."
