Goldman Sachs
Quant Strategist Intern (Summer Analyst)
Key Achievements
- On the Mortgage Strats team, engineered features and trained a Random Forest (scikit-learn) model to predict Commercial Mortgage Backed Securities (CMBS) building valuations.
- On the IRP SMM Strats team, extended the Spreader algorithm in Java to improve performance under aggressive market conditions and investigated execution slippage against arrival market prices, identifying opportunities worth approximately +$137K per $1mm duration traded.
SAIL, KAIST
Research Assistant
Supervised by Prof. Jaesik Choi and Research Prof. Nari Kim
Key Achievements
- Enhanced efficacy by leveraging dependency parser-based syntactic units, resulting in 7% increase in metric scores.
- Evaluated LIME, SHAP, IG, LRP and Attention on text classification models for Plug and Play XAI project.
- Developed a pipeline to provide the best explainer algorithm for particular text-based models to build trustworthiness.
Adobe
Research Intern
Supervised by Kuldeep Kulkarni and Gaurav Sinha
Key Achievements
- Implemented fixed-precision and mixed-precision Quantization-Aware Training algorithms for Generative Adversarial Networks to optimize on model size and performance in PyTorch.
- Analyzed layer-wise sensitivity of StyleGAN2 by estimating average Hessian matrix trace via Hutchinson algorithm.
- Achieved a high compression ratio of about 6.5 times with negligible degradation in FID score as compared to the original model.