Hamid Sabermand

Designed and built production LLM systems using OpenAI API and LangChain, delivering ~94% task success rate with 10K+ daily requests and <3s average response time. Built RAG pipelines over 50K+ enterprise documents using semantic search and an embedding-based document classifier (96% accuracy). Created a prompt-engineering framework for 20+ use cases and reduced AI API costs by 35% via caching, batching, and token optimization. Implemented robust fallback and error handling to achieve 99.5% uptime despite external API instability. Architected FastAPI microservices for 50K+ daily transactions with distributed async workflows for 100+ concurrent tasks, and built a React/Next.js real-time dashboard with improved performance. Optimized database queries by 45% and mentored two junior engineers.

Built FastAPI + LangChain applications with multi-step LLM workflows and semantic search (92% recall, 94% precision), deployed with monitoring for 5K+ monthly requests. Fine-tuned GPT models for domain-specific tasks, reducing hallucinations by 40%. Experimented with Pinecone, Weaviate, and Chroma to optimize RAG performance for different use cases. Delivered 15+ full-stack projects using FastAPI/Django/Laravel and React/Vue.js.

Led backend redesign to support 1M+ daily API requests (peak 15K RPS). Integrated Elasticsearch for large-scale processing (500GB+ daily) and added vector embeddings to enable semantic document retrieval with 94% precision. Implemented Redis caching to reduce database load by 40–60% and improved API latency by 30%. Added resilience patterns including circuit breakers and retries, improving system resilience to 99.8%. Delivered an enterprise web platform with Laravel (backend) and React (frontend), achieving sub-500ms search over 1M+ records. Built production monitoring dashboards tracking 200+ metrics across services.