Saikrishna Poluri

ADM — Finance R2R Lakehouse & Production GenAI Agents

What: I lead end-to-end delivery of Finance Record-to-Report analytics at Archer Daniels Midland. I architected medallion pipelines on Azure Databricks and ADF that integrate JDE, SAP, HFM, IBM DB2, and IBM APGO Web Wire into Delta Lake with lineage and Unity Catalog governance, feeding KPI frameworks for P&L, Budget vs. Actual, Expense Forecasting, EHS, and Capital Allocation. On the same platform I built and shipped three production GenAI agents over PeopleSoft and SAP labor data: a Corporate Headcount & Salaries Data Agent in a Genie Space (SQL shown with every answer for trust), an Employee Policy RAG Assistant on Vector Search and LLM endpoints with cited answers, and a Combined Multi-Agent Assistant with intent routing — hosted as a Databricks App with a custom Streamlit UI, embedded Genie Space, Databricks AI/BI dashboards, and Unity Catalog row- and column-level permissions for HR users.

Impact: Weekly-to-yearly reporting for the Executive Committee and CFO-level consumers, plus governed natural-language self-service over sensitive HR data — all in production today.

Maersk — Fact-Based Reporting & Dremio Semantic Layer

What: I delivered end-to-end data engineering for Maersk's Fact-Based Reporting program: metadata-driven ADF and PySpark ingestion of SAP S/4HANA, SAP AC-DOCA Universal Journal, and Hyperion HFM data into Delta on ADLS Gen2, financial dimension and Universal Journal modeling in Databricks, publication to Azure Synapse Dedicated SQL Pool, and migration of SSAS multidimensional cubes to Azure Analysis Services Tabular. I then implemented a Dremio and Power BI semantic-layer lakehouse and moved managed tables to external tables for governance and cost control.

Impact: A multi-million-dollar annual infrastructure cost reduction from the semantic-layer lakehouse, faster CFO-level dashboards, and lower-latency Finance analyst exploration on Actuals, Plan, and Forecast P&L.

PANYNJ — ICMS Financial Lakehouse & PATH Ridership Migration

What: For the Port Authority of NY & NJ I designed metadata-driven ADF and PySpark ingestion into ADLS Gen2 Bronze for the agency-wide Integrated Cost Management System, integrating SAP ECC FI, IBM Planning Analytics, and Budget PRO to support Plan, Budget, Forecast, and Actuals reporting across Capital and Operating portfolios for Aviation, PATH, and Tunnels, Bridges & Terminals. In parallel I migrated legacy VBA and MS Access PATH ridership workloads to parameterized Azure Synapse pipelines with Power BI dashboards.

Impact: Agency-wide and division-level financial portfolio visibility for Finance leadership and retirement of fragile desktop tooling — all within a five-month contract.

ADM Treasury — “Forecasting the Flow” (Hackathon: 2nd place, Bard’s Tale Award)

What: On an 11-person hackathon team I built the Databricks lakehouse and data/feature pipeline behind a probabilistic 90-day cash-flow forecasting platform for ADM Treasury, and served as Lead Presenter on Demo Day. I ingested 15 Agris ERP tables plus Maxar weather, corn-futures, and USDA feeds into a Bronze/Silver/Gold medallion under Unity Catalog, and engineered the governed feature and treasury-facing views that fed the forecasting models. I contributed to the model design alongside our ML engineers, who built the cascaded quantile-regression models (XGBoost/LightGBM) that turn those features into P10/P50/P90 cash-timing intervals.

Impact: The team placed 2nd of the field and won the Bard’s Tale Award; delivery-timing MAE of 6.5–7.6 days (26–36% better than a naive baseline), with ~80% of outcomes inside the P10–P90 band — giving Treasury planning intervals instead of a single, brittle point estimate.

Microsoft Certified: Fabric Data Engineer Associate

DP-700 — earned.

Microsoft Certified: Fabric Analytics Engineer Associate

DP-600 — earned.

Microsoft Certified: Azure Data Engineer Associate

DP-203 — earned.

Microsoft Certified: Azure Data Fundamentals

DP-900 — earned.

Microsoft Certified: Azure Fundamentals

AZ-900 — earned.

Databricks Lakehouse Fundamentals

Databricks accreditation — earned.

Dremio Verified Lakehouse Associate

Dremio — earned.

Databricks Certified Data Engineer Associate

In progress — planned, target 2026. Not yet earned; this knowledge base is my structured preparation for it.

For a Databricks Data Engineer role

• Production Spark and PySpark at enterprise scale, with Delta Lake design covering CDC, SCD 0/1/2, schema evolution, compaction, and Z-ORDER tuning.
• Real orchestration on the platform: Databricks Jobs (Lakeflow Jobs) coordinated with ADF and Redwood RunMyJobs, gated by Azure DevOps CI/CD with monitoring, runbooks, and hypercare.
• Databricks FinOps: cluster right-sizing and autoscaling, Delta tuning, and the Maersk managed-to-external table shift — an estimated ~$500K in annual compute savings at ADM.
• Closing the gap deliberately: Lakeflow Declarative Pipelines, Auto Loader, COPY INTO, and Asset Bundles are active exam and lab study (Databricks DE Associate → Professional in progress), tracked in this hub's Data Engineer Track.

For a Resident Solutions Architect role

• End-to-end delivery ownership: technical planning, cross-team dependencies, stakeholder alignment, and hypercare for CFO-level Finance R2R analytics.
• Architecture advisory in practice: fit-gap recommendations on batch vs. near-real-time, CDC/SCD design, and Power BI vs. Dremio/Denodo semantic-layer selection.
• Governance over sensitive data: Unity Catalog lineage plus row- and column-level controls protecting HR and finance datasets in production.
• Production GenAI on the Data Intelligence Platform: Genie, Vector Search RAG, multi-agent routing, and Databricks Apps — shipped, governed, and serving 100+ users daily.
• Mentoring and enablement: design reviews, code reviews, and PySpark standards for junior engineers.