I build systems that turn messy real-world data into records a machine can actually use.
Type something messy. Watch it resolve. This runs entirely in your browser.
Currently in healthcare, which has the messiest data there is. The problems generalize.
Matching free-text names to canonical records when the same thing is written six different ways in six systems. Products, companies, addresses, medical tests — same problem everywhere.
RAG pipelines over large reference corpora, including hybrid dense and lexical search, multi-tenant filtering, and verification layers that let a system decline rather than answer wrongly.
Turning scanned documents and unstructured text into structured records — OCR, layout handling, schema-constrained generation, and the validation that keeps bad rows out.
ETL and backend services that move large volumes reliably: queue-driven processing, cloud-native deployment, and the cost engineering that keeps a model-heavy pipeline affordable.
Distributed systems, computer vision, and applied IoT.
Towards an IPFS-based highly scalable blockchain for PEV charging and achieving near super-stability in a V2V environment
Advanced traffic violation control and penalty system using IoT and image processing techniques
Conversion of real images into cartoonized image format using generative adversarial networks
I did not write my first line of code until well into adulthood, and I am the first woman in my family to study abroad. Five years later I build production machine learning systems for a living.
That route shapes how I work. I am good at the unglamorous middle of a system — the parts where data is wrong in ways nobody documented, and where being careful matters more than being clever.