At Frankfurt For Information Technology, I design and ship production AI pipelines combining RAG retrieval, LLM orchestration, and API integrations. These systems automate internal processes and client workflows, cutting manual effort from weeks to hours.
I built connect.raizn.com, a multi-tenant marketing intelligence SaaS with OAuth integrations and AI-generated performance recommendations. I also applied semantic chunk retrieval and adaptive rate limiting to process 10,000 products in hours without manual intervention.
On Kalim AI, I built a bilingual shopping assistant for Shopify grounded in live product catalogs. Its RAG pipeline uses Gemini embeddings and pgvector, with catalog updates synced through Pub/Sub webhooks.
I designed Kalim AI's three-layer conversational memory and split the system into independently deployable Spring Boot, Spring Cloud Gateway, and FastAPI services. I also fixed an async race condition in Spring Boot with targeted updates that prevented data corruption under concurrent load.
