Harish N.

3+ years of work experience

Feb 20267 mosPresent

Bengaluru, Karnataka, India

• Built LLM orchestration and tool-calling framework supporting various LLMs with dynamic model routing based on context limits, latency, and cost optimization, along with reusable context caching to reduce token usage across queries • Engineered microservice-based data processing architecture for medium and large-scale workloads using PySpark, Azure Databricks, and Azure ML, enabling distributed execution of structured and unstructured data pipelines • Developed and optimized end-to-end data workflows including connectors, ingestion systems, parallel dynamic processing, profiling, parquet/CSV transformations, and large-scale dataset summarization for downstream analytics and storage • Migrated core analytics layer from Pandas-based computation to SQL/DuckDB-based execution engine, redesigning aggregation and filtering logic to improve memory efficiency and support scalable query-driven processing without breaking visualization outputs • Built cloud-agnostic data export and workflow execution systems using connector registries and serialized execution modules (.pkl-based loaders), along with robust debugging and stabilization of production pipelines (Snowflake, OneLake, workflow execution, and data previews)