About Me

Home About Me

Harrison Jansma

Hi, I'm Harrison

Machine Learning Engineer & Data Science Manager. I lead Generative AI for Capital One's contact centers.

Common questions

At the start of 2018 I'd just finished a math and business degree at Baylor and had no idea where my career was headed. After some honest reflection, I committed to learning everything I could about machine learning and the tech industry — teaching myself, writing about it, and taking on freelance and research work while I earned a master's in Computer Science from UT Dallas.

Six years later, I'm a Data Science Manager at Capital One, leading Generative AI for our contact centers. My teams build real-time, LLM-powered tools that help thousands of servicing agents mid-conversation — agentic summarization, retrieval-augmented generation over live transcripts, and streaming sentiment and complaint detection. Along the way I've also built the auto-loan underwriting models that decision billions of dollars a year. I care most about taking hard, ambiguous problems from prototype to production, and about the measurable difference that makes for the people on the other end.

My Experience

Common Questions

What are you working on right now?

I lead Generative AI for Capital One's contact centers — real-time, LLM-powered and agentic tools that augment thousands of servicing agents live, mid-conversation. Recent work includes an agentic call-summarization pipeline that cut agent handle time and saved $1.3M a year, retrieval-augmented generation over live transcripts, and an LLM complaint-detection system that raised detection from 30% to 80%.

What's your experience with generative AI and LLMs?

Deep and hands-on. I've built agentic, multi-step LLM reasoning pipelines, retrieval-augmented generation with vector search over live data, and "agent-as-a-judge" evaluation stages that certify quality before anything reaches a user. I work provider-agnostically and care as much about evaluating non-deterministic systems as about building them.

Do you have production and scale experience?

Yes. I've architected parallel real-time inference pipelines sustaining 200+ transactions per second at roughly 100 ms latency on Kafka, AWS Lambda, DynamoDB, S3, and Snowflake, monitoring tens of thousands of live calls a day.

Just as important, I build the observability and governance around these systems — dashboards and alerting in New Relic, PagerDuty, and Splunk — so a non-deterministic system can be trusted in production.

What's your leadership experience?

I manage a team of ML engineers and data scientists, set technical objectives with VP- and EVP-level stakeholders, and lead annual roadmapping that aligns our investment with business priorities. I've also acted as interim product manager on new bets — running pilots, MVP demos, and roadmaps to get systems funded and shipped.

What kind of role are you looking for?

I'm drawn to fast-moving product companies where I can build new things, work alongside people who are stronger than me, and directly improve the experience of the people using the product.

The work I'm proudest of is all customer-facing — systems that changed an outcome in a real moment. I want more surface area to do exactly that, whether the title is applied scientist, ML engineer, or engineering leader.