Kushal Subedi

I am a Data Engineer focused on building production-grade, low-latency data pipelines and self-hosted infrastructure for large-scale blockchain and ML workloads. I design distributed stream-processing systems using Kafka and Flink and optimize OLAP query performance for analytics at scale.

In production roles I have architected Kappa-style pipelines processing terabytes of Solana blockchain data with sub-second latency and high availability, and implemented fault-tolerant streaming, transformation, and enrichment workflows that guarantee exactly-once semantics.

I have a strong MLOps background, deploying LLM-powered applications, model serving stacks, and CI/CD for ML using tools like LangChain, TorchServe, MLflow, and Airflow. I also build graph database solutions for relationship analysis and fraud detection using SurrealDB and Neo4j.

I contribute to developer education through workshops and fellowships, and I pursue applied research in multimodal AI and reinforcement learning, having delivered award-winning GenAI and RL projects while maintaining pragmatic, observability-driven DevOps practices.