vinod0x01
Bengaluru, India

Vinod Patil

Senior Software Engineer (SDE3) at Lowe's — AI agents & search infrastructure

I build the AI agents behind Lowe's MyLow chatbot — the layer that turns "which drill do I need for this?" into an actual product answer. Before that I led Lowes.com's move off Elasticsearch, and off keyword matching.

65%LLM token cost cut
3Mdocuments indexed
2,000peak QPS served
5 yrsbuilding at scale
Work
Senior Software Engineer (SDE3)
2026 — Present
Lowe's
  • Architect and lead the Product Discovery Agent — an Agent2Agent (A2A) service on Google ADK, FastAPI and MCP, with Redis-backed agent memory and MongoDB transcript persistence.
  • Cut LLM token consumption by 65%. Agent cost is the problem nobody mentions until the bill arrives — the fix was a memory-consolidation scheme that periodically compacts what the agent has learned, plus dynamic graph workflows.
  • Shipped AI Overview on Search and Product Listing pages, and Smart Filters natural-language filtering that maps plain-text intent onto Lowe's filter taxonomy.
Software Development Engineer
2024 — 2026
Lowe's
  • Led the Elasticsearch → OpenSearch migration to semantic and hybrid retrieval across a 3M-document catalog serving 500–2,000 QPS at peak.
  • Cut search p95 latency from 50–100 ms to 20–30 ms cluster-side; a custom compression plugin contributed 11 ms of that and reduced index storage 65%.
  • Operated two OpenSearch clusters across two regions — 20 nodes each steady-state, scaled to 35 for Black Friday/Cyber Monday — with zero-downtime blue/green refreshes.
  • Built the Apache Beam/Dataflow indexing pipeline and Airflow DAGs, including a custom OpenSearch sink because no official Beam connector existed.
  • Found a bug in OpenSearch itself affecting semantic search, and worked with the upstream maintainers through to the fix.
  • Mentored and onboarded 6+ engineers across search and agent services.
Product Developer
2021 — 2024
EdgeVerve Systems (Infosys)
  • Backend engineering on Finacle, Infosys' core banking platform — Java/Spring Boot microservices holding key APIs under 300 ms at p95.
  • Automated Kubernetes deployment with Ansible across five targets (on-prem, OpenShift, GCP, AWS, Azure); cut service-mesh resource consumption 30% through Istio tuning.
  • Banking is a good place to learn why reliability is not a feature you add later.
Stack
languages
PythonJavaSQL
ai / agents
Google ADKAgent2Agent (A2A) MCPMulti-agent orchestration RAGPrompt engineering
search
OpenSearchElasticsearch Semantic & hybrid searchVector search EmbeddingsRelevance tuning
backend
FastAPIasyncioSpring Boot RedisMongoDBPostgreSQL KafkaApache BeamAirflow
platform
GCPKubernetes / GKEDocker IstioAnsibleJenkins PrometheusGrafana
Recognition

Team Excellence Award

Lowe's India · 2025

For shipping semantic search on Lowes.com. Read more

Modern Retail Awards — Finalist

Best Use of AI in Retail · 2025

Lowe's semantic search platform, which I helped build, was named a finalist. Read more

Education
M.Sc. Computer ScienceLiverpool John Moores University
2023 — 2025
PG Programme, Software Development (Cloud)IIIT Bangalore & upGrad
2022 — 2023
B.E. Computer Science & EngineeringKLS Gogte Institute of Technology
2015 — 2019