Architecting Resilient Data Pipelines & Agentic Systems.
Modernizing enterprise data lakes with PySpark, Databricks, and Delta Lake across AWS & GCP. Pioneering autonomous agentic data engineering workflows with Claude Code, Antigravity, and Codex.
Forward-Deployed Impact Powered by Autonomous Agents
I am a Forward Deployed Data Engineer at Verizon and a Data Science graduate from the University of Minnesota. I specialize in bridging raw distributed data infrastructure with high-stakes business outcomes.
My development methodology is agentic-native: I prioritize Claude Code as my primary autonomous CLI partner for rapid PySpark transformation scaffolding, test synthesis, and infrastructure automation, supplemented by Google Antigravity and Codex for architectural modeling and deep static validation.
The Agentic Pairing Triad
Query Shiva's Technical Intelligence
Interactive agent terminal modeling Shiva's architecture, forward-deployed experience, and autonomous workflows.
Autonomous Data Pipelines with Claude Code
- Autonomous Pipeline Generation: Utilizing Claude Code to rapidly scaffold PySpark medallion transformation models, automate YAML workflow configs, and orchestrate schema diffing scripts.
- Test Synthesis & Static Verification: Claude Code automates edge-case unit test generation for data reconciliation, validating row-level assertions before execution on clusters.
- Zero-Leak Defense: Claude Code is configured with strict credential avoidance rules—automatically stripping tokens, utilizing
.env.exampletemplates, and validating against localgitleakspre-commit hooks.
The 3 Project Pillars
Structured to showcase autonomous AI engineering, cloud cost-optimized infrastructure micro-tooling, and production-grade data modeling.
AI End-to-End Consumer App
ANTeater (open-source Yuka alternative): Scan a grocery barcode for a transparent 0–100 health score, additive warnings with sources, and a better option in the same aisle. Optional Gemini reads ingredient photos with your own key; no backend, no stored keys.
Cloud & Data Infra Micro-Tools
80 days, two features a day: one Databricks and one Google Cloud build daily, each on a platform feature released in the last six months: pipelines, governance, cost guardrails and ML in SQL. Each one tested, cost-capped and keyless.
Legacy & College Engineering
Machine Learning & Analytics: Unsupervised player clustering on NBA performance data, computer vision ball/strike calculation for slow-pitch softball (SLOW-ARC), and Enterprise Sales Intelligence dashboards in Tableau.