• Implemented multi-object tracking for parcels on industrial conveyor belt systems across multiple camera views, reducing manual security checks from 5 minutes per parcel to real-time automation

• Applied fine-tuned YOLO models and developed 3D pose estimation for parcel detection and localization

• Developed data visualizations for parcel detection and tracking performance analysis

• Optimized forward kinematics computations within a proprietary simulator for a curbside delivery robot

• Streamlined the transition to high-performance computing by refactoring and documenting core simulator modules and overcoming key performance bottlenecks

• Owned the complete development and deployment of a Retrieval-Augmented Generation (RAG)-based solution, enabling publishers to track and monetize proprietary data in LLM applications, leading to a joint venture with a strategic partner and $100k+ in external investment

• Developed a scalable AI-powered platform that enables research groups and e-commerce businesses to access publisher data on an as-needed basis through tailored AI Agents, reducing data-acquisition costs compared to traditional subscription packages

• Thrived in a fast-paced early-stage startup environment, constantly adapting to rapidly changing requirements, adopting new skills, and independently scoping and defining technical tasks

• Optimized system performance by iterating on open-source LLMs (e.g., LLaMA, Mixtral), embedding models (using transformers and the HuggingFace API), RAG techniques (LangChain), and vector databases (chromadb, ElasticSearch) to improve search accuracy and response time

• Built internal tools, rapidly iterating on Streamlit apps, to test hypotheses and automate dataset creation, reducing manual workload and development time

• Streamlined CI/CD workflows with GitHub Actions and Docker, reducing deployment time and improving system reliability through automated unit and performance testing

• Built an LLM-powered language tutor used in a Spanish department pilot program

• Designed and implemented backend logic and the chat storage API, used to interface with SQL databases

• Leveraged open-source LLMs (e.g., LLaMA, Mixtral) using the transformers package and HuggingFace API

• Tailored solutions to classroom needs through close collaboration with faculty

• Instructed students on control algorithms in C to perform autonomous driving tasks

• Assisted students to design and debug hardware and software for a miniature autonomous vehicle

• Assessed performance and behavior of students’ autonomous vehicle systems