Hyderabad, India · Software engineer
Mahesh Vanam
I like the messy middle between AI agents, data pipelines and the screens people actually use. At Gainsight I give agents safe, autonomous access to production analytics, keep 10-million-record pipelines running, and cut a data lake to a sixth of its cost.
Selected work
Agents
Designed an MCP server for custom analytics: agents query KPIs, dashboards, and data-lake metrics as structured tools, with schema validation and scoped permissions keeping every action safe.
Agentic coding workflows — LLM code generation behind automated review gates, shipping higher-quality code faster.
Pipelines
Architected a pub/sub bulk-operations service with decoupled producers and consumers, sustaining 10M+ records/hour through peak load.
Restructured ETL resource allocation and culled redundant cloud infrastructure across multi-level data-lake pipelines — a 6x cut in operational cost.
Built the data-lake foundation for structured and unstructured data at scale, plus 4 self-serve connectors and Python/PySpark/AWS ETL turning raw community data into real-time insight.
Interfaces
Revamped analytics dashboards around 40+ customizable atomic KPIs; built the internal alerting system that cut incident-response latency 10x.
JIRA integration from zero to shipped in one sprint — requirements, architecture, API design, deployment — letting customers create and escalate tickets in-product instead of switching contexts.
Earlier
Peer-to-peer file sharing on IPFS + blockchain with ECC encryption — confidentiality without a central server.
Deep-learning detection of pneumonia and tuberculosis in chest X-rays, from literature review to high-accuracy model.