Abhinandan
Dubey

Prompt compression and token cost optimizer for LLM APIs. Multi-pass entropy scoring, TF-IDF pruning, response caching, and model routing.

Metric interceptor and job watchdog for Spark. Collects query plans, tracks disk spill and shuffle, kills resource-hogging jobs.

Builds graphs from SLIC superpixels for CRF-based image segmentation. Published on PyPI. Born out of cell segmentation research at Stony Brook.

Automated essay scoring using RNNs and feature engineering. Semantic analysis, clause parsing, discourse flow.

Synthetic Minority Over-sampling Technique implemented for SparkML. Handles class imbalance in large-scale distributed datasets.

Harvests openly licensed artwork from the Louvre, the Met, the Art Institute of Chicago, and the Rijksmuseum.

A language-learning app with a dark, editorial aesthetic. Share text from anywhere, extract vocabulary, quiz yourself.

An AI wardrobe app. Scan your clothes, get AI descriptions, generate outfits. Minimal interface, editorial sensibility.

An RNA world simulator. Nucleotides bond, pair, and form compounds in a procedurally generated biome.

Serialize Spark DataFrames to and from Avro. A bridge between two serialization worlds.

Resource manager for JupyterHub. Also a moon of Jupiter.

Few-shot in-context learning for classifying malicious HTTP traffic. Fine-tunes BERT on packet captures from CTU-13 and WRCCDC datasets.