IMC Trading • Incoming Quant Trading Intern June 2027

Chicago, IL

Amazon • Software Engineer Intern June 2026 - September 2026

Seattle, WA

• Architect a distributed ML experiment-orchestration platform (AWS Step Functions) that compiles declarative configs into fault-tolerant DAGs of multi-node GPU workloads, running unattended multi-day pipelines.
• Design an order-independent checkpoint-selection algorithm and parallelize evaluation fan-out to overlap with in-flight training; cut end-to-end runtime by >50% and inter-stage handoff latency from hours to seconds.
• Build a config compiler that deterministically deduplicates state machines, cut experiment setup from ~30 minutes to <1 minute, and streamline research iteration with agentic failure recovery.

UCLA Scalable Analytics Institute • Machine Learning Researcher January 2026 - present

Los Angeles, CA

• Restructure a 3.9 TB PET/CT archive into a longitudinal panel of 80 subjects × 4 timepoints, and evaluate 4 pretrained encoders. Forecast subject-level embeddings one timestep ahead under leave-one-subject-out cross-validation, beating a random-walk benchmark on 66% of held-out transitions while avoiding look-ahead bias.
• Build a vision-language model, injecting scan and forecast embeddings as image tokens into a LoRA-finetuned 8B language model with joint regression and classification heads; predict disease outcome 8 weeks ahead at 0.76 AUC / 72% vs. 0.58 / 50% scan-only under matched-budget ablation (10 LOSO runs). Submitting to ICLR.

Scale AI • Machine Learning Intern January 2025 - January 2026

San Francisco, CA (Remote)

• Built statistical evaluation rubrics and Dockerized agentic RL evaluation environments, supporting industry-standard AI benchmarks (e.g., Aider LLM Leaderboards) across 200+ datapoints and 6 clients.

Perpetual Futures Funding Carry Strategy May 2026 - present

Personal Project

• Built a C++ market-data simulation pipeline processing 400M+ order-book events across 12 months of BTC/SOL data, reconstructing historical books and modeling execution costs and slippage at varying order sizes.
• Designed a decoupled ingestion and simulation architecture using a lock-free queue, enabling high-throughput evaluation of a funding-rate carry strategy; measured 0.8 Sharpe net of costs and $210K executable capacity.

S&P500 Return Forecasting October 2025 - December 2025

ACM AI, Kaggle Competition

• Ensembled daily-horizon forecasting models for S&P500 forward returns using gradient-boosted trees, feed-forward networks, and LSTMs. Achieved 1.48 Sharpe over a 6-month scoring period; top 22% of submissions.