our (collegiate) magnum opus. designed this with three friends as a part of our senior design project, but it turned into something much bigger.
when i moved into my apartment, i was lucky enough to have core furniture already installed by the guy who lived here before me (i just paid him for it). but when i wanted to buy something new, like a lamp or a new couch for the living room, etc., i faced an issue which is pretty common among home decorators:
a fastapi-powered backend that automates the entire cold-emailing process for job applications. it handles scraping job info to generating and sending emails. i connected it to a google sheets app script and it runs locally with an ngrok tunnel to interact with from the browser for the apps script.
you start by pasting a linkedin job link into the sheet and the system kicks off a playwright script that scrapes the company name, job title, and job description. after this a separate playwright worker searches bing for recruiter or hiring manager profiles at the company (e.g. “recruiters at {company_name} site:linkedin.com”), finds the most relevant result, and adds that link to the sheet.
over the course of jan 2024 - march 2025, i worked on improving, building, and ultimately evaluating an epileptic seizure detection algorithm using “DeepSOZ: A Robust Deep Model for Joint Temporal and Spatial Seizure Onset Localization from Multichannel EEG Data” (Shama, et al., MICCAI 2023) as a baseline model.
a big thank you to Dr. Archana Venkataraman and Deeksha Shama for working with me over this year. they taught me so much content-wise (that i never learned in any classes) but more specifically the process of iterating over an end-to-end research project itself.
when i worked with state street corporation as a part of Boston University’s cloud computing class (ENG EC 528), i built a system to automate resource deployments (orchestrating database migrations) from snowflake over the course 4 months.
this project exposed me to the inner workings of ci/cd pipelines and cloud-native deployment, in addition to how database migrations truly work.
this figure shows the overall architecture of our project, detailing the specific functions of liquibase and harness.
