Payment Event Processing Pipeline — Scala Streaming Backend
Scala 3Cats EffectFS2RedpandaPostgreSQLMongoDB
Scala 3 backend for deterministic JSONL replay, Redpanda-backed stream ingestion,
PostgreSQL enrichment, explainable risk decisions, and idempotent MongoDB persistence.
- Built a backend system with JSONL, paced replay, and Redpanda input modes behind one source abstraction.
- Implemented parsing, validation, normalization, customer enrichment, eligibility checks, and deterministic risk scoring.
- Used PostgreSQL for customer profiles and MongoDB for processed transactions, eligibility violations, alerts, and risk history.
- Structured the code around hexagonal architecture, narrow ports, functional streaming, MUnit tests, Docker Compose, and GitHub Actions CI.
Aegis AI — GCP SRE/ChatOps Platform
Google CloudTerraformCloud RunGKEPub/SubBigQueryFirestoreSlackGemini
GCP prototype for cross-project incident detection, Slack alerting,
metric-backed follow-up answers, and auditable incident storage.
- Built a cross-project SRE/ChatOps workflow connecting GKE workload logs, Cloud Logging sinks, Pub/Sub, Cloud Run services, and Slack.
- Used Firestore for incident sessions and BigQuery for incident lifecycle events, reporting views, and SLO evidence.
- Integrated Cloud Monitoring and Gemini so engineers could ask Slack follow-up questions with real incident context and metrics.
- Provisioned Hub and Client infrastructure with separate Terraform stacks, least-privilege IAM, Secret Manager, and documented demo runbooks.
Mamut — AutoML toolkit for tabular classification
PythonPyPIscikit-learnOptunaEnsemblesReports
AutoML workflow for tabular classification: preprocessing, hyperparameter optimization,
model comparison, ensemble search, and generated reports.
- Built preprocessing pipelines for imputation, scaling, encoding, skew correction, outliers, and optional feature reduction.
- Supported model search across common classifiers with Bayesian or grid search.
- Added dynamic ensemble search with hard/soft voting, HTML reports, notebook plots, and optional SHAP.