CML Insights App - Causal ML Platform
WorkTechnologies
Key Highlights
- ▸Principal-level ownership of a production causal ML platform from design to scale
- ▸Designed causal inference and agentic LLM systems for decision support and executive reporting
- ▸Led architecture, feature delivery, and technical direction across the ML team
- ▸Built scalable ML systems on Kubernetes with Terraform/Kustomize IaC
- ▸Set engineering standards through design reviews, mentorship, and documentation
Overview
Principal-led causal machine learning platform that goes beyond prediction to explain why outcomes change and what actions work. Built for higher education institutions to surface key trends in student success metrics, identify high-impact subgroups, and prioritize interventions using statistically validated causal effects.
The platform also generates executive-ready impact reports summarizing the effectiveness of applied interventions, translating complex causal analyses into clear, decision-oriented narratives.
Architecture
Cloud-native, multi-tenant ML platform:
- Data: Normalized PostgreSQL schemas for features and experiments
- Compute: Python microservices for ETL and causal modeling
- Orchestration: Dagster-driven training and retraining workflows
- Infra: Kubernetes on GCP with fully managed IaC
Key Technical Contributions
Causal Inference Engine
Designed and implemented the platform’s causal engine using propensity score–based methods with scalable matching, assumption validation, and performance optimizations for large datasets.
ML Platform & Pipelines
Built automated pipelines ingesting heterogeneous data, generating domain-specific success metrics, training models, and deploying results with monitoring for data quality and drift.
Executive Impact Reporting (Agentic LLM System)
Designed and implemented an LLM-powered reporting system that synthesizes causal analysis outputs into executive-level impact reports. The agentic workflow reasons over treatment effects, subgroup responses, and uncertainty metrics to generate concise summaries, visual insights, and recommendations suitable for leadership and stakeholders.
Bridges technical causal results with business-ready communication, enabling faster, evidence-backed decision-making.
Technical Leadership
Owned system architecture and roadmap execution while leading and mentoring the ML engineering team. Drove design decisions, advised on complex tradeoffs, and ensured scalable, maintainable systems.
Technologies
- Python: Scikit-learn, Pandas, NumPy, Dask
- Kubernetes: Production microservices
- PostgreSQL: Feature and experiment storage
- GCP: GKE, Cloud SQL, Cloud Storage
- MLOps: Dagster, Kubeflow
- IaC: Terraform, Kustomize, ArgoCD
Impact
Enabled data-driven institutions to allocate resources to programs with proven causal impact. Delivered a scalable platform supporting rigorous, real-world causal analysis across multiple student success initiatives.