Sunny C.

I build production multimodal AI systems that turn images, documents, and business data into reliable outputs. My recent work includes reference-based visual reasoning and image-generation pipelines with deterministic answer and uniqueness validation, SVG/PNG rendering, bilingual content, and secure FastAPI/Docker APIs. I also build production AI agents and RAG systems that connect LLMs to private documents, databases, APIs, and enterprise workflows. I own the lifecycle from requirements and architecture through structured model outputs, programmatic validation, deployment, and handover. SELECTED RESULTS • Built and deployed Visual Reasoning AI Studio, a reference-based multimodal AI question-generation and deterministic verification platform with a live synthetic showcase in my Upwork portfolio. • Delivered non-verbal image reconstruction and .NET/API integration for an education workflow. • Built a real-time multi-agent cybersecurity platform that improved incident-response speed by 40%. • Delivered an image-to-data pipeline with 95%+ accuracy across complex document types. • Built natural-language-to-SQL assistants for non-technical teams. WHAT I BUILD • Multimodal AI, computer vision, image processing, and document intelligence • Reference-based image and diagram generation with structured visual outputs • AI agents and multi-agent workflows using LangGraph, LangChain, AutoGen, and LlamaIndex • Production RAG, extraction, classification, and structured-output pipelines • FastAPI, Docker, MCP, API, ERP, and business-workflow integrations • Azure/AWS deployment, evaluation, security, and human-approval controls CORE STACK Python, FastAPI, OpenAI, Azure OpenAI, AWS Bedrock, LangGraph, LangChain, LlamaIndex, AutoGen, MCP, computer vision, OCR, SVG/PNG rendering, PostgreSQL, Docker, Kubernetes, Databricks, and Kafka. WHY CLIENTS HIRE ME • Production systems, not isolated prompt demos • Full-stack AI ownership from architecture through deployment • Deterministic validation for correctness, uniqueness, and output consistency • Clear communication, documented decisions, and realistic estimates If you are building a multimodal AI, computer-vision, visual-reasoning, RAG, or enterprise LLM application, send me the current system and desired outcome. I will propose the smallest reliable path to production.