I build production AI systems, LLM-powered pipelines, multi-agent architectures, and the infrastructure required to make them reliable at scale. My focus is on the full engineering stack: from evaluation frameworks and retrieval systems to deployment, observability, and latency-sensitive inference.

I care about systems that actually work: well-defined failure modes, reproducible evaluation, and clean abstractions between components. Most of my recent work sits at the intersection of applied LLM engineering and data infrastructure; RAG systems, agentic workflows, and the tooling needed to iterate on them without breaking production.

I hold a Bachelor's and Master's in Computer Science (Concentration in Machine Learning). I write about ML systems, evaluation, and engineering decisions at https://thenumbercrunch.com/.

Side projects include HiveHaven — a housing platform for international students in the U.S. — and PolNet, a political network visualization tool for analyzing U.S. congressional caucus data.

Current reading: Build a Large Language Model (From Scratch) by Sebastian Raschka.

LLM Systems, Multi-Agent Architectures, Agentic Workflows
Retrieval-Augmented Generation, Vector Search, Embedding Pipelines
Evaluation Frameworks, Observability, Model Behavior Analysis
Inference Optimization, MLOps, High-Throughput Serving

Python, JavaScript, C/C++, SQL
LangChain, LangGraph, LlamaIndex, Google ADK
PyTorch, Scikit-learn, Hugging Face
Elasticsearch, Neo4j, Postgres, BigQuery
Apache Spark, Databricks, Hadoop (HDFS)
Vertex AI, Vertex AI Agent Builder, Google Cloud Run, GCS
AWS SageMaker, Amazon Bedrock, Azure ML
Docker, Kubernetes, Git, DVC

LinkedIn: https://www.linkedin.com/in/pathak-ash/
X: https://x.com/pathak_jeee
Email: ashutoshpathak@thenumbercrunch.com
Writing: https://thenumbercrunch.com/