Real-Time Artifact Search Engine
Hybrid vector + lexical search scaling to 110k+ items at <50ms latency
At GreatFrontEnd, I designed and deployed a scalable, real-time artifact search engine combining vector (semantic) and lexical search on Railway.
Using a test dataset, I scaled the system to over 110,000 data points while keeping search latencies under 50ms. Alongside this, I led development of an ad and social media layout-generation pipeline using Gemini’s multimodal models, drawing on published research in automated layout generation.
Skills: image segmentation, vector databases, semantic + lexical retrieval, LLM orchestration, OCR, rembg.