Tonna Nwachuku

Engineered a high-performance link extraction engine in TypeScript and Node.js capable of parsing and classifying millions of hyperlinks, including complex cases like multiline or encoded URLs. Implemented a real-time WebSocket workflow synchronization and developed a PDF analysis pipeline with CIDFont-encoded text handling to ensure consistent link repair visualization across PDF types. Built a bulk upload and queued processing system with progress tracking and plan-based limits, and integrated AI services (Perplexity, Anthropic, CommonCrawl, OpenAI) for URL recognition and contextual link reconstruction. Optimized backend with asynchronous tasks, concurrency-safe logic and improved frontend UX with filterable analysis panels and WebSocket-driven state management. Deployed containerized, CI/CD pipelines on AWS EC2 with monitoring for uptime and fast iteration.

Built a global payroll platform and a financial insight and prediction dashboard to forecast cash flow and expenses. Implemented backend APIs in Django and FastAPI with a Next.js frontend. Integrated Hugging Face Transformers for automatic financial report summaries, and LangChain with RAG workflows to generate contextual summaries from structured and unstructured data. Designed and scheduled Airflow ETL pipelines processing over 20 million transactions daily. Deployed services with Docker and Kubernetes on Azure; optimized resource usage to reduce cloud costs by 22%.

Developed AI-driven tools to assist radiologists in reviewing and managing medical images. Built a web-based workflow system using FastAPI (backend) and React (frontend). Integrated OpenAI and CLIP models to summarize radiology reports and align findings with images. Implemented Celery with Redis for background AI inference tasks and introduced Neo4j-based data lineage for traceability of datasets in model training and reporting. Deployed containerized microservices to AWS ECS Fargate with CI/CD automation; reduced inference latency by ~30% using gRPC batching and ONNX.