Market-Based Credit Rating
KMV/Merton credit-rating pipeline: EM asset estimation, PIT/TTC probability of default, and S&P-equivalent ratings for public companies.
- Python
- pandas
- Merton/KMV
I am Jacky Jiang, a B.S. student in Computer Science and Statistics at the University of Illinois Urbana-Champaign (’27). I am also an operator, optimist, and communicator with active listening and strategic thinking. Pace, grit, and empathy are the primary tools I use to achieve high-level results.
In industry, I’ve completed five internships across five different industries — finance, entertainment, logistics, edtech, and agriculture. Across all five, what I enjoy most and do best is helping businesses turn raw data into better decisions. On my own time, I run an agentic trading system and an end-to-end retention analysis covering 50K users and 1.5M streaming events.
In research, I co-authored Steer2Adapt (advised by Prof. Jiaxuan You, UIUC) — inference-time LLM adaptation via composed steering vectors, accepted at the ICLR 2026 Re-Align Workshop and under review at NeurIPS 2026. I also build the real-time diarization and live-summarization pipeline for ScribeAR, an open-source captioning platform for deaf and hard-of-hearing students.
Looking forward, I’m excited to further deepen my AI and data science toolkit expertise, as well as GPU programming with CUDA. Let’s connect to explore collaborations in AI-driven analytics, machine learning, or quantitative trading!
Off the clock: tennis, baseball, and photography.
KMV/Merton credit-rating pipeline: EM asset estimation, PIT/TTC probability of default, and S&P-equivalent ratings for public companies.
Multi-stream GPU edge detection over whole image directories with CUDA NPP and custom kernels — 103 images in 0.64 s on an A100.
Churn & conversion modeling, A/B testing, survival analysis, and uplift-based targeting capturing ~70% of retention value at 38% exposure.
Real-time global map and dashboard of AI-industry news, regulation, and market impact, paired with an LLM benchmarking arena.
ICLR 2026 Workshop on Representational Alignment (Re-Align) · under review at NeurIPS 2026
Composes semantic steering vectors via Bayesian optimization over a prior subspace, improving downstream performance by +8.2% on average across 9 tasks and 3 open-source LLMs — without retraining.
Dean's List · GPA 3.91/4.00
IB Score: 40/45 (Mathematics HL: 7/7)
Ask me anything about internships, research, or full-time opportunities.