Krishna Goje
Freelance AI Data Engineer · Forward Deployed Engineer · 15+ Years

Krishna Goje
I build the data platform,
the AI agents on top,
and own the delivery.

Databricks · Snowflake · AWS · Data Migration · Agentic AI — 15+ years at American Express, Grubhub, Booking.com & Opendoor; currently Forward Deployed Engineer for Turgon AI, with recent work for Siemens Energy. Available for remote contract / B2B work from India: Germany, EU, UK, US and Middle East.

Platforms Shipped. Agents Built. Fraud Stopped.

Not just code shipped — outcomes delivered, from regulated banking data to AI agents on client platforms.

$5M
Ernie AI VoiceBot
Architected the data pipeline powering Opendoor's AI sales agent (Ernie). The bot automates lead outreach with personalized market insights, targeting the 80% of seller leads that human agents never reach — projected to save $5M annually in operational costs.
@ Opendoor
$100M
Customer Network Graph
Led a team of 8 building a Neo4j graph of 150M customers and ~1.5B links across devices, IPs, phones and emails, refreshed daily. It caught bust-out fraud rings before they acted: ~$100M in reduced losses every year.
@ American Express
162
Production Go-Live
Took a leasing-intelligence platform to production on Azure and Databricks, with 162 source tables loading nightly, an LLM chat-with-data assistant, observability and disaster-recovery design. Data ready ~55 minutes earlier each morning.
@ Cenomi · via Turgon AI
6
Agent Migration Pipeline
Built and validated a 6-agent legacy-warehouse-to-cloud migration pipeline: profiling, business-rule extraction, mapping and validation, with confidence-based human escalation and a full audit trail. Caught 100% of seeded defects on client-approved test data.
@ DaVita · via Turgon AI
15+
AI Self-Service Agents
Built a multi-agent system on Claude with 15+ MCP integrations (Snowflake, Databricks, dbt, Airflow, QuickSight, Slack). Stakeholders get answers in seconds instead of waiting half a day to 2 days.
@ Opendoor
ULD
Universal Listing Directory
Built the universal restaurant prospect list: crawled Google Maps outward from seed cities, enriched with competitor menus (DoorDash, Uber Eats) and matched against a 300K+ restaurant platform with entity resolution.
@ Grubhub

15 Years of Building

From engineer to team lead to running my own practice — across banking, marketplaces, travel, real estate, energy and healthcare.

Sep 2026 — Present
Freelance AI Data Engineer · Founder
Varilom Technologies
Remote
  • Forward Deployed Engineer for Turgon AI: client-facing technical lead for Parker Hannifin, Cenomi, DocuSign and DaVita
  • SAP data migration to a Databricks lakehouse; 162-table production go-live; 6-agent migration pipeline
  • Wrote a 161-rule engineering standard, adopted company-wide, plus 9 agent skills that enforce it
  • Built a Snowflake-native platform for Siemens Energy (Snowpark, dbt, Openflow, Cortex)
  • A team that delivers data-platform projects end to end
Aug 2024 — Sep 2026
Staff Data Engineer
Opendoor
Hyderabad, India
  • Owned operations analytics across the iBuying lifecycle: the source of truth for 70%+ of operational decisions
  • Multi-agent AI self-service on Claude with 15+ MCP integrations
  • Architected the data pipeline for the Ernie AI VoiceBot ($5M projected annual savings)
  • Consolidated 100+ datasets into unified models; semantic layer with governed metrics in dbt
Jan 2023 — Aug 2024
Lead Data Engineer
Booking.com
Hyderabad, India
  • Led a 5-person team delivering tax-compliance reporting for every country Booking.com operates in
  • Event-driven CDC lakehouse on AWS: migrated 15+ financial reporting pipelines off legacy MySQL
  • Cut pipeline compute cost 35% under a ~$100K/month AWS budget; cross-region disaster recovery
Nov 2021 — Dec 2022
Lead Data Engineer
Grubhub
Manhattan, NY
  • Built the universal restaurant prospect list (Google Maps crawl, competitor menus, entity resolution vs 300K+ restaurants)
  • Spark on EMR, Kafka, Airflow; mentored 2 engineers on a team of 5
Jan 2017 — Nov 2021
Lead Data Engineer
American Express
Manhattan, NY
  • Named Data SME for Consumer card data; owned ~30 core EDW tables for 100+ applications, with PII governance (GDPR, RBI)
  • Led a team of 8 on a Neo4j customer graph (150M customers, ~1.5B links): ~$100M in reduced losses every year
  • Fraud feature pipeline (2,000+ features from 40+ sources); tripled Spark throughput
  • Hackathon Finalist & Data Geek award winner
Apr 2015 — Jan 2017
Data Engineer
American Express (via Scepter Technologies)
Manhattan, NY
  • Built a real-time federated query engine over HBase, Cassandra and Elasticsearch with one SQL interface
  • Caching cut repeat query time 10x
Aug 2013 — Dec 2014
Graduate Research Assistant
East Texas A&M University
USA
  • Optimized university digital infrastructure

What I Build With

A full-stack data engineering toolkit honed across Fortune 500s and high-growth tech companies.

📊 Analytics & BI
Amazon QuickSight QuickSight Q (NLQ) dbt Semantic Layer Power BI Tableau Looker Data Modeling Data Warehousing Data Governance Data Mesh
🤖 Agentic AI
AI Agents Multi-Agent Orchestration MCP Claude OpenAI LangChain / LangGraph crewAI RAG Human-in-the-Loop Prompt Engineering
☁️ Cloud Platforms
Snowflake Snowflake Cortex AWS Redshift Databricks Unity Catalog Delta Lake Microsoft Fabric Azure Azure Data Factory GCP / BigQuery Terraform S3 EMR Glue Lambda CloudWatch Kinesis
🔄 Data Integration & Migration
Data Migration CDC Apache Airflow Fivetran SAP AWS DMS AWS Step Functions Select.dev Azkaban Oozie
⚡ Big Data
Apache Spark Spark Streaming Hadoop MapReduce HDFS Hive Trino Pig Kafka Neo4j
💻 Languages & More
Python SQL Scala Java Spring Boot Gradle Shell Grafana Elasticsearch Redis Cassandra HBase Kubernetes

Certified & Verified

Industry-recognized certifications across cloud, data, and software engineering.

🧱
Certified Associate Developer for Apache Spark
Databricks
☁️
Solutions Architect - Associate
Amazon Web Services
☕
Certified Professional, Java Programmer (OCJP)
Oracle
🌐
Specialist: HTML5, JavaScript & CSS3
Microsoft

Two Years at Opendoor

Aug 2024 – Sep 2026. Built and owned the operations data platform behind 70%+ of operational decision-making, from lead to resale.

70%+
Of Operational Decisions
6
Executive Dashboards
100+
Datasets Consolidated
8+
Core Data Models Rebuilt
15+
MCP Integrations
2
Leadership Awards
6 Executive Dashboards
Built and enhanced Weekly Operations Review, Sales Performance, Home Lifecycle, Property Assessments, Market Capacity, and C-Suite Executive Summary. Consolidated 3 fragmented reporting tools into unified views, used daily by operations leadership.
Ernie VoiceBot Data Pipeline
Built the data pipeline powering Opendoor's AI sales agent — personally credited by leadership in the company-wide launch announcement. The bot targets the 80% of seller leads that human agents never reach, generating personalized market insights for every call.
End-to-End Operations Data Platform
Architected a unified data platform spanning the entire property lifecycle — from initial home assessment through acquisition, renovation, listing, and final sale. Built an AI-powered executive summary layer on top that auto-generates a 10-line brief for C-suite, surfacing wins, risks, and action items across every stage. Also contributed to accountable.opendoor.com — the company's public investor transparency dashboard.
8+ Data Model Overhauls
Overhauled core data models across the acquisition, inventory, sales, and reporting layers. Fixed long-standing data quality bugs, added financial settlement tracking, renovation type analytics, and mobile app integration. Part of a 17-dashboard, 5-domain data ecosystem.

Recognition from Leadership

🏆
Sr. Director of Operations — Formal company recognition award + public Slack praise: "Another huge win. Thank you."
🏆
Director of Home Operations — Formal recognition award for rapid dashboard delivery and turnaround speed
💪
Director of Data (skip-level manager) — Granted full authority to proactively deprecate outdated metrics: "Just nuke it. Don't ask."
⭐
Operations Analytics Leads — Repeated praise across channels: "Amazing, thank you!" & "Thanks for digging into this... Very helpful."

Trusted by Leaders & Engineers

Recommendations from directors and engineers I worked with at American Express.

A talented developer and engineer with a strong focus on efficiency and optimization — he was a reliable ally throughout numerous complex and technical development workstreams. I always appreciated his kindness and willingness to give his time to others, whether to help with personal technology queries or teach rooms full of people about Spark optimization.
MK
Matthew K. Meyer
Director, Enterprise Fuel
American Express
Krishna is a fantastic person to work with. He is incredibly smart, kind, and has a great intuition. Krishna is great with large datasets and is able to simplify complex problems. I would highly recommend him for your team!
SM
Sarah Mackey
Director of Product, Business Mobile Experiences
American Express
One of the engineers who shows the utmost dedication at what he does. He has in-depth technical and programming knowledge. He is adept at code optimizations and building reusable software components. Very organized and always focuses on the impact his work brings.
YT
Yash Tekena
Data Engineering
Meta (prev. American Express)
Very knowledgeable and hardworking. Not only he completed his work on time, but he also helped in tracking any technical issues which may arise in production. Well versed in Big Data technologies and has excellent problem-solving skills. I would recommend him for any position.
NM
Nikhil Sagar Mekhala
Data Engineer II
Amazon (prev. American Express)
AI-Powered Engineer

AI Agents on Real
Enterprise Data.

At Opendoor I built an agent system that learns from every session and orchestrates 15+ platforms. For clients I now build agents into the delivery itself: migration pipelines, GenAI review layers and agent skills that enforce engineering standards.

Before — Traditional Workflow
⏳ Write SQL manually for each market
⏳ Run queries one at a time
⏳ Click through QuickSight UI for updates
⏳ Manually check dashboards for errors
⏳ Context-switch between 15+ tools
⏳ Repeat mistakes across sessions
Days per feature
→
After — AI-Augmented Workflow
⚡ AI generates SQL from natural language
⚡ 3 agents × 8 queries run in parallel
⚡ Python library updates dashboards programmatically
⚡ Auto-triggered data quality validators
⚡ One interface orchestrates all 15+ platforms
⚡ Self-learning system remembers past mistakes
Hours per feature
2d→secs
Stakeholder Answers
76
Custom AI Skills
15+
MCP Integrations
30+
AI-Managed Projects
4,874
Lines in QS Library
9
Self-Learning Hooks
krishna@opendoor ~ claude-code (example session)
# 1. Data investigation: 3 parallel agents × 8 queries each
$ claude "Why are Phoenix contracts 20% lower this week?"
→ Auto-triggered: data-investigate skill
→ Agent 1: Market volume (PHX, ATL, DAL, HOU, TPA, CLT, RLG, ORL)
→ Agent 2: Funnel drop-off tracing (PHX vs 4-week avg)
→ Agent 3: Upstream pipeline + Airflow DAG health
→ Root cause found in 45s: Pricing model update rejected 34% of PHX offers
# Previously: 2+ hours of manual SQL, tab-switching, Slack searching

# 2. Dashboard update: AI builds the visual end-to-end
$ claude /quicksight "Add renovation cost variance by market to weekly review"
→ Loading automation library (146 functions)... Backing up dashboard...
→ Writing SQL → Testing in data warehouse → Updating dataset → Adding visual
→ Validating data refresh... Verifying visual rendered correctly...
→ Done. Dashboard live with new metric in 90 seconds.

# 3. Slack Data Bot: drafts answers to stakeholder questions
$ claude /slack-data-bot "Check my mentions"
→ Found 3 @mentions from stakeholders in #ops-data
→ Investigating Q1: "What's our close rate by market this week?"
→ Running parallel Snowflake queries... Writer drafting response...
→ Reviewer checking accuracy... Iterating for clarity...
→ Draft ready for review before sending.

# 4. Self-learning: every session makes the next one faster
$ claude /save-session
→ Captured: 3 anti-patterns, 1 speedrun optimization, 2 new error recoveries
→ AI config auto-updated. Next session starts smarter.
⚡ The AI Operating System I Architected

A custom-built ecosystem where AI sits at the center, orchestrating 76 specialized skills, 15+ platform integrations, and a self-learning feedback loop — all tuned to the company's business domain, data schemas, and team workflows.

Claude Code
→
76 Auto-Triggered Skills
→
Parallel Agent Swarms
→
Business Outcomes
Snowflake (Data Warehouse)
QuickSight (BI Platform)
Databricks (ML/Analytics)
dbt (Data Modeling)
Airflow (Orchestration)
Slack MCP
Glean MCP
GitHub MCP
Linear MCP
Select Star
Google Workspace
Gmail
Self-Learning Hooks
→
Anti-Pattern Memory
→
Session Learnings
→
Auto-Config Evolution
Semantic Memory MCP
↔
Persistent Context Beyond Session Limits
↔
Knowledge Carried Across Sessions

What This Enables

Bots, dashboards, pipelines and insights, all driven by agents through one interface, with a human approving what ships.

💬 L1 Support Bot (Slack Data Bot)
Built an AI agent that monitors Slack @mentions, auto-investigates data questions using parallel Snowflake queries, and drafts polished responses via a Writer/Reviewer iteration loop. Acted as L1 data support on behalf of stakeholders — answering questions that used to require an analyst's full attention.
📊 AI-Powered Executive Summaries
Developed Snowflake Cortex-powered natural language summaries embedded directly in QuickSight dashboards. Executives got a 10-line AI-generated brief on how operations are performing — caveats, wins, and action items included — instead of scrolling through dozens of charts.
💻 Dashboard Automation Engine
Built a 4,874-line Python library with 146 functions for programmatic dashboard management. Dashboards were built through AI: SQL generation, dataset updates, visual creation, data refresh — all with auto-backup and destructive change protection.
🧠 Self-Learning System
Every work session feeds learnings back into the system. It captures mistakes to avoid, winning patterns, and domain knowledge automatically. Each new session starts with what the previous ones learned.
"Every engineer will use AI tools. I build the system around them: it learns from its own mistakes, remembers what worked, and keeps a human in charge of what ships."
76 custom skills · 15+ platform integrations · Self-learning feedback loops · Parallel agent swarms · L1 support automation · AI-generated executive insights
Get in Touch

Open Source

Production-born tools extracted from real enterprise workflows and open-sourced for the community.

⚙
quicksight-mcp
MCP Server · Python · Open Source

An AWS QuickSight MCP server — 55 tools for dataset management, analysis editing, visual creation, SPICE orchestration, and dashboard publishing. Built-in self-learning that tracks error patterns and SPICE timing.

55 Tools Safe Writes Auto-Backup Self-Learning
Born from managing 6 executive dashboards at Opendoor. Extracted from a 4,874-line production library used daily.
View on GitHub →
💬
slack-data-bot
MCP Server · Python · Open Source

Autonomous data Q&A bot that monitors Slack for data questions, investigates using parallel AI agents across Snowflake, dbt, and metadata tools, then drafts polished responses via a Writer/Reviewer quality loop. Human-in-the-loop — never auto-sends.

8-Strategy Search Parallel Agents Writer/Reviewer Human-in-Loop
Born from handling 50+ weekly data questions at Opendoor. The bot drafts responses that sound like they came from the engineer, not AI.
View on GitHub →

These tools were built for production. So was I.

LinkedIn

Education

🎓
Master's in Computer Science
East Texas A&M University
Aug 2013 — Dec 2014
GPA: 3.52
📚
Bachelor's in Information Technology
JNTU Hyderabad
Sep 2008 — May 2012
71%

Let's Talk About What You're Building

I'm most useful when the problem is messy, the data is scattered, and the team needs someone who can turn chaos into a system and then put AI agents to work on it. If that sounds familiar, reach out.

Available for remote freelance / B2B contracts from India (Germany, EU, UK, US, Middle East): data platforms, migrations, semantic layers and AI agents. Through Varilom Technologies I can also bring a team for end-to-end delivery.