Building a training dataset from Mars orbital imagery and rover navigation pictures to develop a foundational model for Mars. Focused on Martian image captioning and detection of scientifically significant images, with the longer-term goal of deploying a lightweight model onboard spacecraft to perform autonomous image analysis under severe downlink bandwidth constraints.

Engineering internal LLM infrastructure for a centralized orchestration layer built around LiteLLM, giving teams an OpenAI-compatible API across Gemini, Claude, and GPT model providers. Building a prompt management system with versioned templates, metadata-driven variable injection, REST endpoints, usage guardrails, token validation, rate limiting, cost tracking, and commercial key isolation.

Investigated whether Spiking Neural Networks can align with biological latent manifolds by adapting Joint-Embedding Predictive Architectures to neural population data. Built snnTorch and LeJEPA training pipelines with LIF neurons, spike binning, and Canonical Correlation Analysis metrics to compare artificial neural trajectories against biological activity.

Performing research under Dr. Julie Simons on knowledge gaps in foundational calculus and writing classes, and investigating how LLMs can be used to address them in an equitable way Designing a curriculum that encourages Calculus 1 students to use AI as an effective learning tool in a responsible way Analyzing Student sentiment on the effectiveness of LLMs as a tutor, as well as quantitative performance indicators like test scores and grades

Mentor and resolve conflicts among 60 residents and continuing student residents to foster a supportive and inclusive living environment, collaborating closely within a team of 40 resident assistants and housing staff to coordinate residential programs. Provide 24/7 on-call support to over 2,500 students, promptly addressing emergencies and immediate concerns, while also developing and implementing community-building programs regularly attended by groups of 20 or more students.

Performing research under Dr. Marcella Gomez on machine learning for biological control and wound-healing systems. Currently focused on HealNet, a PyTorch wound-healing model that uses ResNet image embeddings and multi-task heads to classify healing stage and forecast progression day. Earlier work engineered a fully automated Python and CellPose pipeline to segment bacterial spore microscopy images, extract morphology features, and support closed-loop population prediction with 0.3ms processing latency per image — that pipeline is on hold pending new experimental data.

Teach intensive private swimming lessons to children aged 3-12, focusing on water safety and swimming skills, Improved swimming ability from non-swimmer to safe swimmer within 3 hours of lessons. Managed scheduling and communication with parents, ensuring a positive experience for both children and parents. Developed personalized lesson plans to meet the unique needs of each child, resulting in significant progress in swimming skills and confidence.

Previous Roles: Junior Lifeguard (summer 2020) and Lifeguard (summer 2021) Teach swimming to groups of 6–10 children across all 5 Red Cross levels, substantially improving swimming and water safety skills in well over 1000 students Supervise swimmers at a public pool with an average of 200+ swimmers per day, ensuring safety in swimmers Apply first aid and basic life support skills during emergencies, including CPR, Oxygen Administration and wound care, maintaining composure in high-stress environments Collaborate and effectively communicate with a team in a high-stress environments to ensure swimmer safety and successful resuscitation