Intensity Driven. Data Driven Decisions.
AI Systems Engineer · LLM Applications · Backend Platforms
I build AI applications and the backend systems behind them: retrieval, APIs, data models, authorization, and asynchronous workflows. I work with Python, React/TypeScript, and PostgreSQL, with a focus on testable behavior and recoverable failures.
I’m curious about how systems work, where they break, and what makes them dependable.
Founding AI Engineer at Sylox. I carry work from requirements and system design through implementation, review, testing, and release.
Have a time to go through my portfolio:
| Project | What to explore |
|---|---|
| Invoice Followup | Free browser-based invoice follow-up queue and editable payment reminders, with CSV import/export. No account or automatic sending. Source |
| CSV Reporting Example | Reproducible Python example that checks two fictional CSV inputs and produces a six-metric report with a calculation receipt. Synthetic data; no live integration or customer results. Source |
| Customer Support System | A Next.js application, Hono API, PostgreSQL data layer, and separate FastAPI service for retrieval, streaming chat, and tool workflows. Fresh setup and checks |
| ContextDock | A native macOS developer tool built with Swift, SwiftUI, and AppKit for thought capture, plan selection, CLI-agent launches, local events, and Obsidian notes. Build and tests |
| Portfolio | React and Three.js presentation of public work, with source links, technical writing, and static profile exports. |
| Fraud Detection | FastAPI scoring prototype with Celery explanations, PostgreSQL, Redis, and a browser demo. Hosted tests and container checks. Separate historical benchmarks do not establish the packaged model’s accuracy. |
My upstream changes focus on correctness, resource use, compatibility, and security.
| Project | Accepted change |
|---|---|
| Prefect | Allow Azure Blob result storage to overwrite on rewrite. |
| dlt | Read Parquet row groups individually to bound ADBC driver memory use. |
| VictoriaMetrics | Stop PromQL integrate() from extrapolating past the end of a series. |
| Unsloth | Hide non-matching chat threads in search. |
| Langfuse Python SDK | Support wrapt 2.x. |
| Tracecat | Add _source field filtering to Elastic Security detection queries. |
| lollms-webui | Harden path sanitization against traversal and add regression tests. |
Open-source record: accepted, open, and closed submissions · Current merged PRs
- Backend and data: Python, SQL, FastAPI, SQLAlchemy, PostgreSQL, Alembic, Redis, Celery, Parquet.
- AI applications: RAG, LangGraph, LangChain, embeddings, reranking, retrieval evaluation.
- Applications and delivery: React, TypeScript, Docker, GitHub Actions, pytest, OpenTelemetry.
-
GitHub achievements — badges awarded by GitHub for platform activity.
-
LeetCode — algorithm and SQL practice.
-
Kaggle — ML notebooks and competition experiments, including the LLM preference CPU baseline.
-
Credly — credential verification, including Astronomer’s Apache Airflow 3 Fundamentals and DAG Authoring records.
Achievement and credential record · Public profiles and technical writing
I’m interested in AI systems and backend engineering roles where ownership includes design, implementation, and verification.