I build agentic-AI systems that do real work in enterprise finance — MCP-based agents orchestrated with LangGraph, monitored with LangSmith, and kept honest by human-in-the-loop review.
They stand on eight years of less glamorous but essential work: large-scale distributed PySpark pipelines, anomaly detection, and applied machine learning. I care most about making enterprise data — structured or not — reliable and usable, whether the consumer is a dashboard, a model, or an agent.
0+
Years in data
0+
Reports analyzed & consolidated
0%
Anomaly-detection precision
02
Capabilities
GenAI & LLM Engineering
Agentic AI workflowsLangGraphModel Context Protocol (MCP)LangSmith · LLMOpsRetrieval over enterprise knowledgeCursor · Windsurf
Agentic AI for enterprise finance — built carefully, with humans in the loop.
Built and shipped an agentic reconciliation system: LangGraph handles the stateful, multi-step orchestration; agents scan financial data for discrepancies and classify issues (business / data / timing) using historical JIRA context over MCP — an analyst reviews every recommendation before it goes anywhere.
Published reusable MCP agents through our internal developer portal, instrumented with LangSmith for tracing, cost/latency monitoring and evaluation — catching prompt and retrieval regressions before production.
Programmatically analyzed 600+ Qlik reports to map dependencies and redundancies, then consolidated them onto modernized data-lake tables — validated end-to-end against general-ledger source of truth.
Built Python validation frameworks and moved report codebases to Git — version-controlled queries, real code review, faster QA with less manual reconciliation.
Oct 2021 — Jul 2025
Data Analytics Engineer
Apple · Cupertino, CA
Large-scale data processing, anomaly detection and ML-driven investigation on petabyte-scale pipelines.
Engineered distributed PySpark / Spark SQL pipelines over HDFS — 30% faster query execution at petabyte scale.
Automated Spotfire dashboards for leadership — campaign ROI tracking that improved effectiveness 10%.
May 2017 — Jun 2018
Student Research Assistant
Missouri University of Science & Technology · Rolla, MO
Used K-Means to identify distinct user-experience states (Frustration, Flow, Boredom) from quantitative UX data; classified them with SVM, Random Forest, mlogit and kNN — evaluated with AUC, Kappa and hypothesis testing. Data wrangling in R (tidyverse, dplyr).
Nov 2014 — Apr 2015
Project Engineer
Wipro Technologies · Hyderabad, India
Built SSIS ETL processes across multiple sources and SSRS reports for stakeholders; advanced SQL (stored procedures, parameterized drill-downs) plus schema design with indexes and constraints.
04
Selected Work
CASE 01 · INTUIT
Agentic AI Reconciliation System
Agents that scan enterprise financial data for discrepancies, classify the issue using historical JIRA context over MCP, and propose a resolution — with an analyst approving every step. LangGraph for orchestration, LangSmith for tracing and evaluation. Less false-alarm triage, faster incident resolution.
LangGraphMCPLangSmithPythonHuman-in-the-loop
Live
In production
CASE 02 · INTUIT
Enterprise Report Consolidation
Programmatic analysis of a sprawling Qlik estate — extracting metadata and table-level dependencies, finding redundancies via SQL analysis, and consolidating everything onto modernized data-lake tables with end-to-end validation against the general ledger.
PythonSQLData-lake migrationAutomated validation
600+
Reports analyzed
CASE 03 · APPLE
Anomaly Detection Framework
Statistical methods (Grubbs' test) applied to flag suspicious user activity at production scale — strengthening data security and pipeline reliability without drowning the team in false positives.
PythonStatistical ModelingScikit-learnSQL
90%
Precision
05
Education & Credentials
M.S. — Information Science & Technology
Business Analytics & Data Science · Missouri University of Science & Technology
2017 — 2018
Machine Learning Specialization
Stanford University · Coursera
Certified
Microsoft SQL Developer
Microsoft
Certified
Alteryx Core
Alteryx
Certified
06
Current Focus
What I'm working with (and thinking about) day to day right now:
Agentic workflows that earn trust — HITL by defaultMCP servers for enterprise dataLLMOps — tracing & evals with LangSmithRetrieval over messy enterprise knowledgeAI-assisted development · Cursor · Windsurf
07
Contact
Always happy to talk data platforms, agents — or the messy middle where they meet.