Deep dives into the engineering decisions, trade-offs, and patterns behind the most complex projects in this portfolio.
Traditional fraud detection systems are rigid — they either rely on static thresholds that generate excessive false positives, or require expensive ML pipelines that are slow to update. RiskPulse needed a detection system that business analysts could configure without engineering involvement, while still supporting complex composite rules.
Coffee recommendation systems typically either use basic filtering (too simple) or LLM-powered suggestions (expensive, unpredictable latency, non-deterministic). Coffee Sommelier needed recommendations that feel personalized while remaining fast, predictable, and cost-free to operate at scale.
Mind Mirror needs real-time sentiment analysis for every journal entry, but the Hugging Face Inference API has rate limits, occasional downtime, and costs that scale with usage. The system needed to be resilient against API failures while maintaining acceptable sentiment quality.