Personal work
Products, open-source work, and labs that keep the engineering practical.
Separate from my professional experience, these projects show hands-on product thinking, architecture, delivery, and the habit of turning ideas into working systems.
Personal product for pay-per-use career tools: CV analysis, ATS scoring, cover letters, interview preparation, LinkedIn optimization, job comparison, and salary insights.
Engineering focus: grounded job intake and source trust, authentication, credit and AI-budget controls, rate limiting, durable AI workflows, persisted results, recovery/refunds, retention/redaction, and turning model output into actionable job decisions.
Personal product built around practical image workflows for real estate agents and e-commerce sellers, plus the Maya conversational assistant.
What it includes: virtual staging, day-to-dusk conversion, background removal, AI headshots, product photography, photo restoration, HD upscaling, and a streaming AI assistant.
Engineering focus: monorepo and multi-app architecture, shared cross-subdomain identity, credit and usage controls, AI image pipelines, streaming conversational AI, rate limiting, conversation persistence, and model/GPU cost controls.
Provider-neutral evaluation framework that combines standard RAG quality checks with principal-aware required/forbidden evidence and deterministic CI release gates. RAGAssure 0.1.0 alpha is released on PyPI.
Engineering focus: executable golden datasets, identity-aware evidence contracts, retrieval precision/recall, latency and cost thresholds, reports, adapters, Docker, PyPI packaging, and release automation.
Public reference application that makes the evidence authorization boundary explicit before generation and uses RAGAssure to catch principal-specific retrieval regressions before release.
Current milestone: FastAPI + Next.js, synthetic employee/manager principals, deterministic candidate retrieval, authorization before generation, allowed-only citations, Docker, and a live RAGAssure CI gate. Next: pgvector semantic/hybrid retrieval, grounded model generation, telemetry, real identity, and stronger policy/data enforcement.
Public Home Assistant App that packages the RustDesk ID/rendezvous and relay services so users can run remote-desktop infrastructure they control.
Engineering focus: Home Assistant App packaging, native amd64/aarch64 container builds, upstream checksum verification, security scanning, GHCR multi-architecture publishing, release automation, and practical self-hosting ergonomics.
Reference n8n workflows for AI content drafting, approval buffers, social posting, before/after showcases, and portfolio-style promotion.
Engineering focus: n8n workflow design, Gemini prompting, webhook patterns, social API glue, Google Sheets review queues, and safe credential templating.
Public proof of concept integrating Azure AI Foundry with Microsoft Power Platform to categorize and prioritize IT asset requests.
Tech: Azure Functions, Azure AI Foundry, Power Automate, Dataverse, Power Pages, Power Apps, and OpenAPI.