I build robust backend systems and deploy AI models that don't just work in a notebook, but scale reliably in production. Bridging the gap between conceptual AI and enterprise infrastructure.
Many AI projects fail to leave the development phase due to inadequate infrastructure and poor backend architecture. I specialize in crossing that chasm, ensuring your models are resilient, cost-effective, and integrated seamlessly into your existing systems.
Designing cloud-native backends that handle high throughput and dynamic loads without failing.
Taking models from raw code to robust, versioned APIs ready for enterprise consumption.
Developed advanced RAG pipelines using Qdrant and LangGraph to process complex technical documents and signage, migrating to an agentic workflow that improved retrieval accuracy and reduced token usage by 15%.
Engineered high-speed SharePoint ingestion pipelines utilizing Temporal orchestration and Microsoft Graph API, achieving 30–40% faster document processing for real-time AI systems.
Built an end-to-end vision pipeline integrating SAM2 for rapid image annotation and Detectron2 for precise segmentation to automate the generation of analytical reports.
Contributed to an automated pipeline for road surface analysis and 3D image processing using computer vision techniques.
Deployed a highly scalable, low-latency cloud pipeline for real-time multilingual communication, combining FasterWhisper, NLLB-200, and xTTS-v2 for seamless speech-to-text, translation, and voice cloning.
Developed a property-query chatbot with persistent conversational memory, containerized with Docker and hosted on cloud infrastructure.
Designed a comprehensive IoT monitoring and automated climate control system using ESP32 microcontrollers, supported by a Node.js backend and a Flutter mobile dashboard.