All work
Founder & Founding Engineer, WENES-AI
WENES CRM: AI-Powered CRM for the Uzbekistan Market
Jul 2026 - Present · MVP in testing
Founding engineer on WENES CRM: I took an open-source CRM core, built on it, and localized it for the Uzbekistan market, integrating AI to semi-automate the workflow from conversations to sales, aiming to help SMEs increase efficiency and revenue. The MVP is now in testing.
Highlights
- Built a RAG pipeline for multilingual documents (including tables and code-switched text): hybrid retrieval (embeddings, SQL search, reranking) plus online and offline evaluation, improving Recall@1 to 0.82 and MRR@10 to 0.9 from baseline.
- Implemented a self-learning system that enriches the knowledge base over time via FAQ, knowledge, and performance gap detection.
- Built a temporal-memory knowledge graph (Graphiti) to track user context over time, plus an agent tool-use loop for reading and writing to local databases and external CRM / ERP systems.
- Custom agent orchestration in Rails: every turn runs a decision layer (scope, language, clarification, handoff), retrieval, the reply, and a judge that checks the reply against its sources before it is sent, with real human handoff.
