I build and ship features on Audible's Android and iOS apps, used by millions of listeners. Off the clock I like the systems behind the screen: data pipelines, recommendation models and LLM tooling.
Shipping native Android and iOS at scale, plus the backend and data work behind it.
Pipelines, models, graphics and hardware, mostly built at NYU and NMIMS.
A model I built with my mates at NYU. Data pipelines over the Million Playlist Dataset, using MinIO and Docker to simulate a cloud-native setup, process lyrics into positive training pairs for personalized discovery across languages.
Uses the Qwen LLM to detect anomalies in network traffic. An end-to-end pipeline turns raw traces into structured corpora for the model.
An autonomous agent with a React front end, built on Hugging Face models and a vector DB, that automates multi-step user workflows.
PySpark analysis of NYC subway ridership from 2020 to 2024: anomaly detection, station clustering and predictive models on MTA hourly data.
github.com/roncell/MetroScan ↗A gesture-based lock on the STM32F429I Discovery board that records motion sequences from the IMU and recognizes them to authenticate.
github.com/roncell/Embedillo-Sentry ↗A 3D cubic Bézier curve editor. Chain segments, drag control points in 3D and animate an object along the path at constant speed.