- Part-time: Jan 2022 - May 2023, Mar 2024 - May 2024- Full-time: June 2023 - Aug 2023, May 2024 - September 2024- working in Bruno Olshausen’s Lab (Computer Vision and Computational Neuroscience Lab) under Anthony Thomas (https://thomas9t.github.io), in the past worked under Denis Kleyko (linkedin.com/in/denis-kleyko-85107657) and Paxon Frady (https://www.linkedin.com/in/paxon-frady-02368b1b/)- Implementing Resonator Networks (RNN built with high dimensional computing) for tense translation task- Built upon, refactored, and wrote codebase for Resonator Networks; designed and wrote testing scheme- Performed literature review and high level design of research project goals, outline, and timeline

- working under Eric Wallace (https://www.ericswallace.com), PhD student in The Berkeley NLP Group Lab (http://nlp.cs.berkeley.edu) on Large Language Model Code Generation Project- Created pre and post processing methods for lab built Codex model (LLM for code generation); goal was to create general set of pre and post processing methods to improve LLMs for code generation.- Ran tests seeing distribution of code generation outputs to design system for correcting inaccurate outputs- Understood Codex model architecture and brainstormed architectural improvements to model performance- Designed prompt reformulation scheme to edit inputs for improved output code generation of LLM- Presented research at BEARS (Berkeley EECS Annual Research Symposium)

- Previous External Committee Member and one of four Recruiting Leads- Completed intensive, four month machine learning bootcamp

- Intern on Machine Learning Team- Worked on creating pipeline to relabel images for eventual retraining of object classification models - Ran tests for application-side servers to see how many vision models can fit on single server; Intent is to reduce deployed servers- Wrote unit tests for internal communication scripts

- Developed data processing pipeline and performed data analysis on user metric data to gauge how FocusCalm product is affecting users; FocusCalm is an app/EEG recording headband that trains users to improve focus and calm abilities through engaging in sessions with neurofeedback training- Created algorithm that calculates EEG frequency band data from raw EEG data collected from FocusCalm SDK- Performed data analysis on FocusCalm EEG frequency band data to see whether a feature visualizing frequency band data should be incorporated into the FocusCalm app- Worked on creating signal processing/machine learning algorithm that measures level of a user’s cognitive state; performed literature review to decide cognitive state that algorithm will measure, and created data collection paradigm to collect data that signal processing/machine learning algorithm is trained on- Utilized: Python, Matplotlib, Scipy, Numpy, Pandas, Signal Processing Techniques, Statistics Techniques

- Developed fully automated Heart Rate, Galvanic Skin Response, and EEG data processing pipelines using python and matlab; all three pipelines are used in several lab studies to generate statistics about study effectiveness- Programmed algorithms for and implemented low-pass, high-pass, drift-removal, and moving average filters for removing artifacts and noise from the signal- Front end developer on team that designed cognitive-training video-game, Meditrain. Limited back end development. Developed with Unity software, using C#. Video-game is currently being used in lab study to improve attention.- Utilized: Python, Matplotlib, Scipy, Mne, Numpy, Pandas, Signal Processing Techniques, Matlab, Unity, C#

- Learnt fundamentals of neurotechnology; machine learning, signal processing, and mathematical techniques to analyze and process fMRI and EEG data. Developed capstone fMRI data processing pipeline (https://github.com/123453ANSH/fMRIAnalysis)

Spent COVID-19 working with an aspiring First Robotics Team; I virtually educated 6 members on basic robotic mechanical and electrical principles, while teaching them C and C++ programming.

- Head/Secondary instructor for classes of 10-30 students, ages 4-52- Private lesson instructor- Third in command at Chois charity events and festivals - Train prospective Chois employees- Lead staff organizer for yearly student picnic and Halloween Party- Form, board-breaking, and sparring tournament referee- Experience working as school manager/human resources: enrolling students and tracking progress/enrollment contracts, updating families about Chois events, selling equipment, giving tours

- Led the 1st Irvington team in 9 yrs to qualify for states & Create US Open Nationals, place 24th at Create US Open Nationals- Won Design award (given to the team with best building process and most innovative robot design) alongside team- Designed circuits, build, and code- Managed tournament schedule, finances, equipment, and progress of four competition robotics teams - Hosted mock competitions for 8 robotics teams to prepare for tournaments

- Created 68 page engineering curriculum - Formed partnerships with two-day cares to host the camp at two locations - Formed partnership with Coding for All Nonprofit for computer supply - Advertised at 3 fairs, 2 schools, and online- Trained 8 instructors- Camp taught 26 students - Raised $2455 in profits through camp. Profits went to rehosting this camp, and funding low-income robotics teams to compete.