Javed Habib
Backend, AI & infrastructure
Software engineer across backend, AI and infrastructure, with 2+ years owning production systems end to end as a founding engineer on a remote team: LLM-agent backends, Kubernetes compute platforms, auth and IAM, and observability, in Python and Rust. Open source first: 2× Google Summer of Code contributor and 2026 GSoC mentor.
Bengaluru, India[email protected]github.com/jaeaeichlinkedin.com/in/javed-habib
looking for the pdf?
Javed-Habib-CV.pdf is the one-page version I send with applications, so it’s shorter and worded a little differently from this page, which is the longer web version with links to the work.
experience
- Built a production agentic AI system (Python, Google ADK, Vertex AI) coordinating 10+ MCP-connected services, with live queries over a 500 GB+ internal database and retrieval across TB-scale public archives, cutting analyst report turnaround from hours to under 15 minutes.
- Architected the Kubernetes-native workflow platform for HIPAA and EU AI Act scoped clients, bringing typical pipeline runs from hours to about 15 minutes across 30+ client pipelines; shipped through ArgoCD GitOps with push-to-merge releases under 5 minutes.
- Designed the authentication and authorisation layer: Keycloak OIDC, a custom IAM service and OPA/Rego policies across 10+ services; hardened the cluster with mTLS, Kubernetes RBAC, NetworkPolicies and Sigstore-signed images.
- Built a chunked, resumable S3 upload service for multi-TB datasets with constant-memory streaming, plus a metadata index serving sub-50 ms search across 5M+ objects; an OCI-runtime sandbox isolates untrusted user code.
- Instrumented a polyglot Python, JavaScript and Rust service fleet with OpenTelemetry, unified in SigNoz and Grafana for end-to-end request tracing.
- Authored the organisation’s Python service template (29 merged PRs): Poetry, type checks, dev containers, complexity gates and hardened GitHub Actions, standardising new ELIXIR Cloud services.
- Owned CI/CD and release-on-merge pipelines across an 8-service fleet; maintained proWES and proTES (GA4GH WES/TES proxy middleware) with type-safety refactors, CI repair and Kubernetes manifest fixes.
- 2024: led the ground-up redesign of TESK, a Kubernetes-native GA4GH TES server in production at universities in Czechia and Greece: Java 8 to Python/FastAPI, image about 10× smaller (650 MB to 70 MB), cold start about 6× faster (45 s to 7 s); added TES data models, a Kubernetes client layer, Helm charts, multi-arch CI and docs.
- 2023: built framework-agnostic Web Components for GA4GH WES and TES on a design-token system; they became the core of the ELIXIR Cloud Components suite (31 merged PRs).
open-source projects
Kubernetes-native GA4GH TES server with a production Helm chart; forked and run in production by an external organisation.
GA4GH WES workflow engine; load-tested at about 2k submissions/s with 10k+ workflows in flight.
S3 client with Python and Node bindings; up to 3× aws-cli download throughput at a fraction of the memory.
skills
- Languages
- Python, Rust, Go, TypeScript, SQL, Bash, Java
- Backend & data
- FastAPI, axum/tokio, REST/OpenAPI, gRPC, Postgres, Valkey/Redis, NATS JetStream, S3/MinIO
- Cloud & infra
- Kubernetes, Helm, Docker, ArgoCD/GitOps, GitHub Actions, Terraform, OpenTelemetry, Grafana, SigNoz
- Security
- Keycloak (OIDC), OPA/Rego, JWT, mTLS, Kubernetes RBAC, NetworkPolicy, Sigstore
- AI
- LLM agents, Google ADK, Vertex AI, MCP servers, RAG
education & leadership
- Google Developer Student Clubs lead and OpenLake open-source mentor: ran workshops on cloud, AI/ML and open source. Upstream fixes to the GA4GH TES spec and the CNCF Harbor CLI.
writing
Design write-ups and postmortems at jaeaeich.com/writing, including the poiesis v0.2 redesign and building and benchmarking s3z.