AI/ML Engineer in New York. I build production LLM systems, RAG pipelines and agentic AI for regulated domains: clinical AI, IP intelligence and legal NLP.
- Currently: AI/ML Engineer at Karyon.Bio, working on the Liver Digital Twin and multimodal HCC detection (genomics + EHR + MRI/CT with MedGemma and TabNet)
- Previously: founding AI/ML engineer at Squirrel IP (Artha, an AI patent valuation engine) and data engineer at Go Digital
- Published: Large-Language Memorization During the Classification of U.S. Supreme Court Cases (arXiv, 2025)
- M.S. Computer Science, Pace University
馃寪 Portfolio: djcodesstuff.github.io 路 馃捈 LinkedIn 路 鉁夛笍 info.joshidhruv@gmail.com
| Project | What it is |
|---|---|
| Spring_2025_project | Research code for my arXiv paper: benchmarking BERT, Legal-BERT, LLaMA 3 and DeepSeek on 279-category Supreme Court case classification |
| LeaseLens | Multi-agent assistant for commercial real estate leasing: Gemini, a lease knowledge graph, Qdrant RAG and Flask |
| StreamLit_app | Rosa, a multi-agent women's health chatbot (Gemini, RAG, FAISS) |
| UAVLogViewer | An AI chat I added to ArduPilot's flight-log viewer that answers questions from drone telemetry (LangGraph, Gemini) |
| HandSign | Hand sign recognition API: MediaPipe landmarks and a Keras classifier behind Flask |
| SAM-NET | Brain tumor MRI segmentation (BraTS 2020): U-Net pseudo-labels train a decoder on a frozen Segment Anything encoder |
| Partial-Data-Clustering | Research prototype chaining DBSCAN and hierarchical clustering, toward classifying incomplete data |
AI/ML: LLMs 路 RAG 路 LangChain 路 LangGraph 路 PyTorch 路 HuggingFace 路 LoRA/PEFT 路 multimodal models Backend & MLOps: Python 路 FastAPI 路 Docker 路 Kubernetes 路 Airflow 路 MLflow 路 Qdrant 路 Redis Data & Cloud: Spark 路 SQL 路 Snowflake 路 AWS 路 GCP Vertex AI 路 Azure