kumi_da

Location: Bengaluru, India Remote: Yes (flexible on EU/Gulf/US-overlap hours); on-site/hybrid also works Willing to relocate: Yes - UAE, KSA, EU, UK, ANZ, SG. Available at short notice

Technologies: Python, Java, Spring Boot, FastAPI | LangGraph, LangSmith, MCP, RAG, NL2SQL, MLflow, Kubeflow | Kubernetes, Helm, Terraform, ArgoCD, Istio, Kyverno, NVIDIA GPU Operator, Volcano | MySQL, PostgreSQL/PgVector, MongoDB, ClickHouse | OpenSearch, Elasticsearch, Lucene | Prometheus, Grafana, Thanos, GitHub Actions Résumé/CV: https://kumida.xyz/resume GitHub: https://github.com/kumiDa Email: [email protected]

Senior Software Engineer / Tech Lead, 7+ years at Visa. I've built and run each layer of an AI platform in production: the agent layer, the retrieval and data systems under it, and the Kubernetes and GPU infrastructure under that.

Now: leading a production LangGraph multi-agent system for enterprise analysts - orchestration, hybrid BM25 + pgvector retrieval, NL2SQL with a validation layer over generated SQL, ingestion from 15+ sources, and a held-out eval set scored every release. I own architecture and interface contracts across several engineering pods.

Before: self-managed Kubernetes across 900+ nodes (networking, policy, GitOps), GPU node pools with Volcano scheduling for the internal ML platform, an org-wide OpenSearch cluster, and self-service CI/CD and secrets platforms adopted by 50+ teams.

I lean toward reliability over novelty: deterministic paths where they suffice, observability before it's needed, build-vs-buy decided on operational cost. Open source: contributions to CPython, pytest and Trino; co-organiser of BangPypers.

Looking for senior/staff roles in AI platform, agentic systems, or platform/infrastructure engineering.

Location: Bengaluru, India

Remote: Yes; on-site/hybrid also works

Willing to relocate: Yes - UAE, EU, UK, ANZ, SG

Technologies: Python, Java, Spring Boot, FastAPI | LangGraph, LangSmith, MCP, RAG, NL2SQL, MLflow, Kubeflow | Kubernetes, Helm, Terraform, ArgoCD, Istio, Kyverno | MySQL, PostgreSQL/PgVector, MongoDB, ClickHouse | OpenSearch, Elasticsearch, Lucene | Prometheus, Grafana, Thanos, GitHub Actions

Résumé/CV: https://kumida.xyz/resume

Email: [email protected]

Senior Software Engineer and Technical Lead, 7+ years at Visa, covering the full stack of an AI platform: the agent layer, the data and retrieval systems beneath it, and the Kubernetes and delivery infrastructure beneath that. I have built and operated each layer in production, an uncommon combination in a field where most practitioners have depth in one.

I currently lead a production LangGraph multi-agent system serving enterprise analysts - agent orchestration, retrieval architecture, natural-language querying with a validation layer over generated SQL, ingestion across 15+ sources, and trace-level evaluation of retrieval quality. Previously: self-managed Kubernetes across 900+ nodes with its networking, policy, GitOps and scheduling stack; an organisation-wide OpenSearch cluster; and self-service CI/CD and secrets platforms adopted by 50+ teams.

My judgement runs toward reliability over novelty - deterministic paths where they suffice, observability instrumented before it is needed, build-versus-buy decided on operational cost. As a lead I own architecture and interface contracts across several engineering pods, and invest heavily in mentoring. Community: contributions to CPython, pytest and Trino; co-organiser of BangPypers.

Seeking senior or staff roles in AI platform engineering, agentic systems, or platform/infrastructure engineering, where platform work is treated as a product discipline.

To a new programmer like me starting my career the comment section in here is an eyeopener.