Abu Bakar Siddik

Leading Core RAG & AI Team • Scaling AI & Backend Systems

Built production-ready RAG system without vendor lock-in, handling real-world workloads

Designed multi-agent workflows with state management, improving throughput by 30%

Developed advanced agentic RAG system with multiple memory layers, increasing contextual accuracy by 45%

Implemented time-aware RAG system with temporal context understanding for recency-based retrieval

Developed web search integration with deep answer feature combining multiple sources and fact-checking

Implemented full-fledged citation system with proper referencing of files, web sources, and memory

Reduced hallucinations by 98% using hybrid search and source-grounding

Built internal evaluation benchmark, cutting evaluation time by 99%

Implemented backend using Hexagonal Architecture (Port & Adapter)

Designed modular ports for OpenAI, Anthropic, and local models

Developed comprehensive LLM evaluation pipeline with source-based fact-checking and validation

Reduced jail-breaking and AI safety risks by 45% with multi-layer guardrails

Implemented multiple end-to-end features: database schema, APIs, frontend, and integration testing

Deployed private AI solutions with on-premise air-gapped operation

Practiced Test Driven Development (TDD) methodology