Siddharth Raipal

Machine Learning Engineer

5-year toxic relationship with machines. They never learn.

Skills

ML / AI

PyTorch PyTorch Geometric scikit-learn GNNs LLMs RAG Knowledge Graphs

Data & Infrastructure

PySpark Databricks Polars Neo4j Pinecone AWS Docker

Frameworks

LlamaIndex LangChain NetworkX Streamlit

About

Machine Learning Engineer with experience building production ML systems and applied research
Previously MLE at Rise11 — built a medical retrieval engine using LLM agents and Neo4j knowledge graphs, with PySpark pipelines processing millions of nodes
At USC, working with InfoLab on clinical interpretable tools for dementia risk prediction
Benchmarking multi-agent RL for robotics with RESL
Phi Kappa Phi scholar
Box with USC Trojan Boxing Club

Research

GlaucomaGuide

Knowledge-Based Systems (KBS) — Under Review

A multi-turn RAG chatbot for glaucoma patient education. Outperformed ChatGPT-4 (+14%) and Gemini (+7%) with 75% task success on complex clinical scenarios.

Clinical Interpretable Tools for Dementia Risk

InfoLab, USC

Building clinical interpretable tools for dementia risk prediction using UK Biobank data.

Multi-Agent RL for Robotics

RESL Lab, USC

Benchmarking multi-agent reinforcement learning algorithms for non-prehensile robotic manipulation.

Projects

Graph RAG Nutrition Chatbot

6-stage Graph RAG pipeline converting unstructured nutrition research into a queryable knowledge graph. Reduced knowledge-graph fragmentation by 60% using graph-enriched entity embeddings.

View on GitHub

ICGRNN

GNN-RNN architecture with end-to-end convexity guarantees for heat diffusion control. Achieved 78% on Cora, 73% on PubMed, and 0.000101 MSE on physics-informed control tasks.

View on GitHub