# SoYeon Ahn

**AI Evaluator / Subject Matter Expert** — United States

> Technical evaluator designing frontier-model evaluations and scoring rubrics at Mercor Intelligence, with a PhD in Electrical and Computer Engineering and expertise in probabilistic modeling and computational biology.

Career Archetype: **Launchpad Navigator** (HS-DG) — One of the first Launchpad Navigators on Saywise

## Links

- LinkedIn: https://linkedin.com/in/soyeon-ahn-64a31266
- GitHub: https://github.com/SoYeonA

## Experience

### AI Evaluator / Subject Matter Expert, Mercor Intelligence (2026-02 – 2026-07)
Remote · Design and validate multi-file data science evaluations for frontier AI models, developing input artifacts, reference answers, and weighted scoring rubrics. Analyze repeated model trajectories to identify systematic reasoning, visual-interpretation, and error-propagation failures. Design adversarial STEM benchmark problems to probe reasoning limits across six frontier models.

### Founder, BlueDove (2024-02 – 2025-07)
Remote · Built and launched a generative-AI commerce MVP solo using React, Node.js, FastAPI, Docker, SQL, and AWS S3 for custom product design and ordering. Owned development and delivery end to end, fulfilling real customer orders concierge-style before winding down in July 2025.

### Project Manager, H/A Studio (2022-10 – 2023-06)
Remote · Coordinated website projects across clients, developers, and designers. Translated founder requirements into actionable tasks and managed timelines and deliverables.

### Freelance Developer, Self-Employed (2022-01 – 2022-09)
Remote · Designed and deployed a full-stack QR-code-based contactless ordering system for restaurants, handling both frontend and backend implementation.

### Postdoctoral Researcher, University of Southern California (2018-05 – 2020-09)
Los Angeles, CA · Built automated Python/Snakemake pipelines integrating long- and short-read sequencing data, separating true variants from artifacts. Designed repeat-aware algorithms for de novo human centromere assembly from long-read data.

### Graduate Research Assistant, The University of Texas at Austin (2012-08 – 2018-05)
Austin, TX · Developed aBayesQR (Bayesian inference) and TenSQR (tensor factorization) methods for viral quasispecies reconstruction from NGS data. Developed a Sequential Monte Carlo algorithm for joint haplotype assembly and genotype calling, improving phasing accuracy and reducing runtime on large-scale HPC datasets.

### System Engineer, Hanwha Thales (formerly Samsung Thales) (2010-01 – 2011-02)
South Korea · Developed and simulated radar signal-processing algorithms for long-range target detection.

## Education

### Doctor of Philosophy (Ph.D.) in Electrical and Computer Engineering, The University of Texas at Austin (2012-09 – 2018-06)
Advisor: Prof. Haris Vikalo

### Master’s Degree in Signal Processing(Detection and Estimation), Electrical Engineering, Korea Advanced Institute of Science and Technology (2008-09 – 2010-06)
Advisor: Prof. Joohwan Chun

### Bachelor’s Degree in Electrical Engineering, Korea University (2003-09 – 2008-06)

## Skills

Benchmark Design, Reinforcement Learning Human Feedback (RLHF), AI Evaluation, Computational Workflows, Data Pipeline Development, Modeling and Simulation, Signal Processing, Algorithm Development, Statistical Modeling, Full-Stack Development, Generative AI, Data Science, Data Intelligence, Large Language Models (LLM), React, Node.js, Matlab, Research, Machine Learning, Algorithms, R, Python, Task Design and Validation, Reference Answers, Rubric Design, Model Failure Analysis, Adversarial STEM Benchmarks, Data Quality Review, C/C++, SQL, Git, Linux/Unix, HPC/Cluster Computing, Snakemake, FastAPI, Docker, AWS S3, Bayesian Inference, Probabilistic Modeling, Sequential Monte Carlo, Tensor Factorization, Data Analysis, Computational Biology, NGS Data Analysis

## Credentials

- **aBayesQR: A Bayesian method for reconstruction of viral populations characterized by low diversity**
- **Viral quasispecies reconstruction via tensor factorization with successive read removal**
- **Joint haplotype assembly and genotype calling via sequential Monte Carlo algorithm**

---
Source: https://saywise.com/member2181 (last modified 2026-10-07T18:49:24.792Z)