Confirm the cause with evidence
I connect traces, logs, and database evidence instead of treating the reported symptom as the answer, then carry the fix through prevention.
LEE TEIBACKEND · FULL-STACK · APPLIED AI
Design. Verify. Operate.
I make complex systems
keep working.
I connect data, APIs, interfaces, and infrastructure with a backend-first perspective. My work spans national public systems, AI SaaS, and applied LLM products—both building from zero and recovering what already runs.
I'll show the problem and the evidence before the title.
- 100k+
- daily batch records
- 8 mo.
- incident root cause found
- 209%
- cost KPI achieved
- 3
- new products built
HOW I WORK
What I do well, in three sentences.
Take zero to one all the way
I connect APIs and data models to the interface, deployment, and operations. Done means usable in the real world, not merely demoable.
Design for cost and operations
Reliability, observability, and infrastructure cost are product concerns from the beginning—not chores left until after launch.
SELECTED EVIDENCE
Results you can examine.
01OPERATIONS
Found the cause of an incident unresolved for eight months.
I led the introduction of APM and correlated traces, logs, and database evidence to isolate a connection leak behind an unresponsive API. We checked nine similar locations and documented preventive controls.
8-month incident closed · 9 similar patterns audited02SCALE
Operated systems processing over 100,000 records a day.
I designed and built high-volume batch logic, database procedures, and integrations for citizen-facing public systems where correctness and continuity were non-negotiable.
100k+ records/day · 8 years 8 months in public systems03ZERO TO ONE
Turn ideas into products that operate.
I led architecture and core implementation for three new AI SaaS products. In my own work, I also shipped and operated LLM agent APIs, RAG pipelines, and Kubernetes infrastructure.
3 new products · Applied LLM through infrastructureA SMALL PRINCIPLE
New technology matters less than whether the result works
and can be operated again tomorrow.
I verify what I do not know, and leave important decisions in executable tests, logs, and documents a teammate can read.