Background

Incoming software engineering role in IBM's AI organization. I am keeping the public description intentionally high-level until the internship begins.

Research role focused on deep learning fundamentals, bilevel optimization, and reproducible large-scale experimentation. The work sits between theory, implementation, and publication-quality empirical evaluation.

• Bilevel optimization: Implemented gradient-based principal-agent contract design methods with implicit differentiation, Hessian-vector products, conjugate gradient solves, and stabilization for nested optimization loops.
• Experiment systems: Built reproducible training and evaluation pipelines with standardized configs, seed control, sweep automation, logging, aggregation, and figure generation.
• Deep learning: Worked on self-supervised learning pipelines, including SimCLR and MoCo-style experiments in PyTorch with GPU training workflows.
• Publication work: Contributed to experiment design, result analysis, manuscript writing, and code/figure preparation for academic submission.
• Result: Co-authored paper accepted at NeurIPS 2025 workshop; extended version submitted to ICLR 2026.

Self-directed ML product work around maritime routing, operational risk, and explainable decision support. The project combines applied ML, optimization, geospatial data, and product UI.

• Routing engine: Designed a NetworkX-based routing system with dynamic weights for time, fuel, weather, cost, and risk.
• NLP risk signals: Integrated Transformer-based maritime news analysis to convert unstructured information into route-level risk signals.
• Decision support: Built route comparisons for fastest, cheapest, and safest paths, with explanations so trade-offs were visible rather than hidden.
• Product interface: Developed a Streamlit app for interactive route exploration, constraint changes, and scenario comparison.
• Operations: Added monitoring concepts with Prometheus/Grafana-style service visibility for runtime behavior.

Software development role supporting civil and environmental engineering research through simulation tooling and reusable model components.

• Simulation systems: Built Java and AnyLogic models for infrastructure and environmental process analysis.
• Modeling approach: Implemented agent-based and discrete-event simulations to test performance under realistic constraints.
• Research support: Translated researcher requirements into modular simulation logic that could be reused and adjusted across experiments.
• Collaboration: Worked with graduate researchers to debug model assumptions, validate outputs, and refine scenario parameters.

Data and probabilistic modeling role focused on uncertainty, resilience, and supply-chain risk. This work connected statistics, modeling, and applied engineering decisions.

• Probabilistic modeling: Applied Bayesian Networks to represent supply-chain dependencies and reason about resilience under uncertainty.
• Inference: Implemented MCMC and inverse modeling workflows to calibrate probabilistic models from data.
• Analysis pipeline: Built Python and R workflows for cleaning, exploratory analysis, visualization, and reporting.
• Communication: Converted model outputs into interpretable research deliverables for technical audiences.

Early research experience in quantum communication and secure protocol modeling, which shaped my interest in the intersection of physics, computation, and uncertainty.

• Security modeling: Studied parity-qubit communication and analyzed leakage, errors, and eavesdropping resistance.
• Protocol design: Modeled secure transmission protocols and explored quantum coin-flip style problems.
• Tooling: Used Qiskit and computational experiments to reason about secure communication behavior.

Community-focused leadership role coordinating a remote intern group and supporting outreach work.

• Led a team of 15 interns; coordinated weekly standups, delegated work, and tracked deliverables.
• Produced social content and reports to support outreach and communicate ongoing initiatives.