Gunjan Kaphle — Senior Engineering Manager
~/career
~/career$

Tab completes · recalls · Ctrl+L clears · the document below follows along

Resume — plain text

Download PDF ↓

Engineering leader with 10+ years building and scaling Data & ML platforms at high-growth startups and large enterprises. Deep expertise in real-time data infrastructure, production ML systems, and using LLMs to unlock self-serve analytics across an organization. Washington Metropolitan Area.

About

10+ years in data and ML infrastructure at Cerner, The RealReal, AWS, and now GOAT, where I lead the Data & ML Platform org.

Most of that time has gone into systems other teams depend on and rarely think about: streaming pipelines, ML serving, data catalogs, orchestration. Lately, AI agents that handle self-debugging and data requests directly.

Experience

Senior Engineering Manager, Data & ML Platform — GOAT Group

Lead Data Engineer → EM, Data Platform → EM, Data & ML Platform → Senior EM

  • Own end-to-end strategy and architecture across real-time streaming, ML infrastructure, and AI systems — powering ML pricing, real-time inventory, and self-serve analytics.
  • Built a real-time streaming pipeline from scratch on Debezium, Kafka, Kafka Connect, and Materialize, cutting data delivery latency from 2 hours to seconds.
  • Architecting AI agents that fulfill data requests, enforce access controls, detect PII, and reason through pipeline failures toward automated remediation — time-to-data from weeks to hours.
  • Built ML infrastructure on SageMaker hosting 50+ production models; time-to-deploy from weeks to days.
  • Designed a custom feature store on Snowflake and DynamoDB, and a Medallion lakehouse on Apache Iceberg with Spark.
  • Deployed Airbyte and OpenMetadata on EKS, indexing 5,000+ datasets across 10+ systems into a central catalog.
  • Introduced Airflow as the company orchestration layer — 3,000+ production jobs. Cut Snowflake spend 30%.

Software Development Engineer — Amazon, AWS SageMaker

  • Built features across the model deployment lifecycle for SageMaker Hosting, serving millions of real-time inference requests daily, in Go and Java.
  • Designed the pre-production testing framework that became the architectural foundation for blue/green model deployments across Hosting.
  • Built an error-handling framework across Hosting services, improving error clarity and reducing incident MTTR.
  • Automated weekly operational metrics across all APIs, eliminating 4+ hours of manual on-call work per week.

Data Engineer — The RealReal

  • Architected an ETL framework on Python, Kafka, and Airflow feeding BigQuery, with anomaly detection catching data quality issues before the warehouse.
  • Built a cross-service extraction and deletion system for GDPR and CCPA compliance.
  • Established CI/CD on GitHub Actions with tests, coverage enforcement, and linting.

Associate Senior Software Engineer — Cerner (now Oracle Health)

  • Extended Airflow into a Workflow Management System used across teams for ETL and batch jobs.
  • Built a Java REST ingestion API processing 50K files and 100GB daily.
  • Baselined user-experience metrics at 1-minute granularity with Kafka and Vertica.
  • Built a Python Flask app on Ansible deploying to 70K+ hosts.

B.S. Computer Information Systems — Minnesota State University, Mankato

GPA 3.90 · Dean's List, Aug 2012 – May 2015

Selected work

real-time-cdcDebezium → Kafka → Materialize. Replaced a two-hour batch lag with seconds. Feature store hydration, live catalog sync to Google Shopping, and warehouse optimization all fell out of it.
data-agentsAI agents that fulfill data requests, enforce access control, detect PII across datasets, and reason through pipeline failures toward automated remediation.
ml-servingML serving infrastructure hosting 50+ production models across CPU and GPU — both batch and real-time inference.
nl-analyticsLLMs wired into the internal analytics stack so non-technical users ask questions in plain English instead of filing a ticket.

How I lead

Platform work is measured in other teams' velocity.If my org ships a lot and nobody else moves faster, we built the wrong thing.
Eliminate toil.Is your request for more people really a request for less manual work?
Build teams around people who keep evolving.The best engineers stay curious. They learn new technologies, challenge their assumptions, and adapt as the industry evolves. I value people who aren't defined by what they know today, but by how quickly and thoughtfully they can grow into what's next.
Optimize for the speed of decisions.Slow data creates slow decisions. Slow decisions create slow organizations. Good engineering isn't just about making systems faster. It's about shortening the distance between signal, decision, and action.

How I think

Design for the second reader.Someone will inherit this code and this error/alert at 3am without you.
Self-serve is the only thing that scales.Any workflow that routes through an engineer eventually becomes a queue.
Boring technology is a feature.Platforms should be boring. Every novel component spends an on-call budget you cannot see until you are paying it.
Make the failure mode obvious.A pipeline that fails loudly is kinder than one that quietly serves stale rows.

Skills

Leading
platform strategy · org scaling · hiring & mentoring · cross-team roadmaps
Building
real-time streaming · lakehouse architecture · ML infrastructure · agentic systems
Languages
Python · Go · Java · SQL
Data
Kafka · Spark · Flink · Iceberg · Airflow · dbt · Materialize · Debezium · Airbyte · Pinot
Cloud
AWS · Kubernetes · GCP · Docker
Stores
Snowflake · PostgreSQL · DynamoDB · Redis · BigQuery · OpenSearch
AI/ML
LLMs · AI agents · RAG · SageMaker

Writing

I write about data platforms, streaming, and ML infrastructure at medium.com/@gunjankaphle.

Interests

Technical
agentic AI · DuckDB · Apache Spark
Otherwise
traveling · cooking · mountain biking · soccer · basketball