Kubeflow Pipelines (CNCF Graduated)
Enterprise-grade Machine Learning Workflow Orchestration
-
Problem: Large execution manifests caused node-level memory exhaustion (OOM panics) during database queries under high concurrency.
-
Implementation: Engineered backend payload defenses within the Go API server by overriding the SQL query builder to inject mocked schemas, dynamically stripping multi-megabyte payloads.
-
Result: Bypassed expensive JSON allocations during
ListRunsdeserialization, actively mitigating OOM panics while maintaining REST API contract compatibility. -
Problem: Vulnerabilities in the metrics parsing infrastructure exposed the system to denial-of-service (DoS) vectors.
-
Implementation: Architected security mitigations targeting the extraction processes.
-
Result: Successfully patched tarball traversal exploits and zip bomb vulnerabilities.
View Architecture Trade-offs
Implementation Note: Tracing nil pointer panics across the
api_converter.golayer required strict handling of the mocked payload schema. Trade-off: Increased code complexity in the data access layer was necessary to ensure predictable heap memory behavior during large-scale execution retrieval.
Kubeflow Katib
Automated Machine Learning Infrastructure
- Problem: Validation pipeline instability affected hyperparameter tuning controllers.
- Implementation: Enhanced controller unit test coverage and resolved validation logic errors.
- Result: Improved controller stability across the validation pipeline.
Selected repositories: CURA · SFORA · Cyberia · AI Resume Builder
Languages
Frontend & UI
Backend & APIs
Databases & Brokers
Infrastructure
Machine Learning
[2026]Software Engineer Intern Role Certification — HackerRank[2026]Open Source Contributor — GSSoC (DevPath)[2026]Published Indian Patent Application — VIT Bhopal University[2025]Google Cloud Generative AI Certification[2025]Applied Machine Learning in Python — University of Michigan (Coursera)

