Avatar

Stephen Hwang

              

Resume

Bioblurb

Stephen is a Partner Solutions Engineer at DataHub, where he helps design cross-platform data context layers—column-level lineage, observability, and governance—that support agentic workloads for strategic technology and cloud partners. Before DataHub, he spent years designing and deploying AI/ML systems: agentic pipelines in FIPS-compliant environments at Research Triangle Institute, high-throughput FastAPI backends scaling to 3M+ daily transactions, and hybrid full-text/vector retrieval systems that boosted search relevance by 80%. Earlier still, he was a Solutions Architect at MongoDB and a backend engineer building event-driven data pipelines for clinical and genomics research. He's passionate about building AI-driven systems that are reliable, scalable, and built with privacy and security in mind.

Interests

  • Applied AI/ML & agentic systems
  • Data engineering
  • Cloud computing
  • Partner & solutions engineering

Education

  • MSc Bioinformatics

    Johns Hopkins University, Baltimore, MD

  • BSc Bioinformatics

    Loyola University, Chicago, IL

Experience

  • Partner Solutions Engineer — DataHub, San Francisco, CA (May 2026 – present)

    Technical point of contact for strategic technology & cloud partners; designed a cross-platform column-level lineage, observability, and governance layer supporting agentic workloads.

  • Senior Solutions Engineer, Applied AI — Research Triangle Institute, RTP, NC (May 2023 – May 2026)

    Deployed LLM-powered agents in FIPS-compliant environments; engineered a FastAPI backend scaling to 3M+ daily transactions at sub-100ms latency; boosted search relevance 80% with a hybrid full-text + vector retrieval system.

  • Solutions Architect, Enterprise Presales — MongoDB, New York, NY (Jan 2022 – Jul 2023)

    Generated $1.7M in net new consumption across healthcare, financial services, and agriculture manufacturing clients; designed sharded, geolocation-based deployments supporting 20M weekly transactions.

  • Senior Backend Engineer, Cloud Computing — Research Triangle Institute, RTP, NC (Jun 2015 – Jan 2022)

    Reduced data latency 80% with an event-driven ELT workflow (Kafka + Airflow); led cross-institutional analyst and engineering teams as technical lead.

  • Lead Backend Engineer — Oncotect, Raleigh, NC (Jan 2021 – Jan 2022)

    Designed an event-driven, distributed pipeline to aggregate and prepare clinical data for stakeholder consumption.

  • Backend Engineer — Johns Hopkins Medical Institutions, Baltimore, MD (Oct 2011 – Jun 2015)

    Data engineering lead on the Stand Up 2 Cancer Epigenetics Dream Team (su2c.org).

Certifications

AWS Certified Solutions Architect (Associate) MongoDB Database Administrator MongoDB Developer (Associate)

Skills

Programming languages Python, MQL, SQL, R, JavaScript, Perl
Databases MongoDB, Redis, SQL, Postgres, Pinecone, Qdrant, Elasticsearch, Redshift, BigQuery
Data Engineering Airflow, Kafka, Kinesis, Glue, Pub/Sub, SNS/SQS, Dataproc/Spark
AI tooling & frameworks Spec-driven development with agent harnesses, Bedrock, Foundry, Gemini, HuggingFace, LangChain
Data sharing & visualization FastAPI, Flask, Google Looker, Power BI, Tableau, ggplot, matplotlib
DevOps & infrastructure Git, Subversion, Vercel, Docker, Kubernetes, Terraform, CloudFormation, Prometheus

Portfolio

A few things built for fun and for keeps — side projects, experiments, and tools.

Featured

Kei Tracker

Live web app

Tracking the trans-Pacific journey of a Sambar — a real-time tracker built to follow a kei truck's shipment from Japan to the U.S.

Python Vercel Real-time tracking
Visit site Source
Featured

Suhocare

Side project · augmenting human care with AI

Intelligent care agents delivering real-time wellness monitoring, companionship, and fraud protection for loved ones — peace of mind powered by real-time intelligence.

AI agents Care tech
Visit site

MCP + A2A Template

An end-to-end tutorial showcasing a complete agentic pipeline using Anthropic's Model Context Protocol and Google's Agent2Agent for plug-and-play, multi-agent coordination.

MCP A2A Agents
Source

Hybrid Search (Qdrant)

A hybrid retrieval experiment combining full-text and dense vector search with Qdrant — the same approach that boosted search relevance 80% in production.

Qdrant Vector search
Source

Unraid + Grafana/Prometheus

Homelab observability stack — Prometheus metrics and Grafana dashboards monitoring an Unraid server.

Prometheus Grafana
Source

Published work

Vaz, M., Hwang, S. Y., ... Baylin, S. B. (2017). Chronic cigarette smoke-induced epigenomic changes precede sensitization of bronchial epithelial cells to single-step transformation by KRAS mutations. Cancer Cell, 32, 360–376. doi:10.1016/j.ccell.2017.08.006 [PubMed]

  • Designed and created novel visualizations for publication and exploratory analysis, including unsupervised clustering techniques used for samply quality control.
  • Plot types include density plots, scatterplots, heatmaps, bar & line graphs.

Gern, J. E., Jackson, D. J., Lemanske, R. F., Seroogy, C. M., Tachinardi, U., Craven, M., Hwang, S. Y., ... Bacharier, L. B. (2019). The Children's Respiratory and Environmental Workgroup (CREW) Birth Cohort Consortium: Design, methods, and study population. Respiratory Research, 20(1), 115. doi:10.1186/s12931-019-1088-9 [PubMed][Full PDF]

Contact