The Challenge: Building a privacy-first employee monitoring system that ensures compliance while respecting data sovereignty—all activity data remains local, yet the server stays informed for reporting and analytics.

The Engineering: Architected a Hybrid Store-and-Forward synchronization engine using Rust and SQLite.

• Optimized Network Strategy: Implemented a split-protocol where high-volume telemetry is micro-batched to reduce overhead, while critical state transitions (e.g., "Shift Start") trigger immediate, event-driven propagation.
• Fault Tolerance: Designed a local persistence queue that buffers data during outages. The system utilizes an exponential backoff algorithm to retry failed syncs, ensuring eventual consistency without overwhelming the server upon reconnection.
• Data Integrity: Enforced idempotency via client-generated UUIDs to prevent duplicate entries during retry storms.

Stack: Rust, Tauri, Svelte 5, Python, Golang.

The Challenge: Analyzing sentiment from thousands of financial news headlines in near real-time using heavy NLP models on low-cost CPU infrastructure without latency spikes.

The Engineering: Optimized a DistilRoBERTa transformer model using ONNX Runtime and Dynamic Quantization. Reduced server costs to near-zero while maintaining 94% sentiment accuracy. Implemented Celery for distributed task processing with 15-minute intervals to handle high-volume news feeds while managing CPU load efficiently.

Stack: Python, ONNX Runtime, Celery.

The Challenge: Detecting infrastructure outages from major service providers (Cloudflare, AWS, GCP) the moment they occur, then automatically notifying affected clients based on their dependency stack.

The Engineering: Built a real-time web scraper that monitors status pages of major cloud providers. Implemented a PostgreSQL-backed dependency mapping system using foreign keys to track which clients rely on which services. When an outage is detected, the system leverages LLMs to draft contextual notification and apology emails tailored to each affected client's specific service dependencies.

Stack: Python, SvelteKit, PostgreSQL, LLMs.