Khanh Van.
Resume

Career record

From idea to production.

Backend systems, reliable releases, and applied AI.

View résumé
01 / INDUSTRY

Built in production.

Select a role to open the details.

State Farm

Software Engineer II

Enterprise policy APIs, cloud modernization, and AI tools for engineering knowledge.Mar 2026 – Present
  • Built policy APIs handling ~1M requests per day.
  • Architected a multi-source RAG system for context-aware answers across internal engineering documentation.
  • Led the first-place team in State Farm’s 2026 Hackathon with an AI collision-reconstruction app selected for productization.
  • Cut deployment time by ~25% during AWS modernization and reduced cross-domain request latency by up to 150 ms.
  • Java
  • Go
  • Python
  • AWS
  • PostgreSQL
  • Kubernetes
Paycom

Software Developer II

Production scheduling, attendance exceptions, and release automation.Jun 2025 – Mar 2026
  • Reduced scheduling bugs and errors by ~30% through stronger validation.
  • Expanded automated workflow tests and release confidence.
  • Automated builds, tests, and deployments with GitLab CI/CD.
  • PHP
  • C#
  • React
  • MySQL
  • GitLab CI/CD
Alkami Technology

Software Engineer Intern

Backend engineering for real-time fraud detection and alerts.Jun 2024 – Sep 2024
  • Built a rule-based, ML-augmented service with over 90% detection accuracy.
  • Connected PostgreSQL and REST APIs to fraud-alert workflows.
  • Java
  • Python
  • scikit-learn
  • Spring Boot
  • PostgreSQL
  • AWS
Royal Dutch Shell

Retail IT Intern

Location-aware retail features and backend integration.Jun 2019 – Aug 2020
  • Built a geofenced arrival feature with Google Maps for real-time station notifications.
  • Contributed to REST APIs and helped reduce deployment delays by 12%.
  • C++
  • Google Maps API
  • REST APIs
  • ServiceNow
02 / RESEARCH

Questions worth testing.

The University of Texas at Austin

Hallucination Analysis in LLMs

Built a four-model benchmark and trained a response-level hallucination detector.Jan 2026 – May 2026
  • Collected 892 responses across easy, hard, trap, and adversarial prompts.
  • Reached 93.3% accuracy on 179 held-out responses within the study distribution.
  • Python
  • PyTorch
  • Transformers
  • scikit-learn
Study overview Read report
The University of Texas at Dallas

Research Associate

Annotated multimodal meme data for dataset construction and AI research.Oct 2024 – May 2025
03 / EDUCATION

Foundations.

Expected May 2027

The University of Texas at Austin

M.S. · Computer Science

May 2025

The University of Texas at Dallas

B.S. · Collegium V Honors · Computer Science

See it in context

Explore the implementation.

Browse projects