Abstract
In today’s increasingly digital world, automation plays a growing role in how large corporations operate on a day-to-day basis. From talent outreach to final hiring decisions, software tools now sit at the center of processes that were once entirely human-led. While automation promises gains in efficiency (especially at scale), it also raises serious questions about the ethics of delegating consequential decisions to code. This thesis portfolio explores these tensions through two projects that, although different in scope and implementation, each grapple with how automated systems shape professional outcomes. The technical project focuses on building a web scraping tool for Paraform, a startup that automates the collection of executive data for cold outreach. The STS research paper examines how AI-powered recruiting tools, when left unchecked, can encode and amplify bias in hiring decisions. Together, these projects offer insight into how automated systems can be designed and deployed responsibly.
The technical project centers on the development of a backend scraping tool for the growth team of Paraform - a startup that streamlines talent sourcing for early-stage companies. The tool was designed to automate the process of identifying top-level executives - such as CEOs and talent partners. This information was scraped by using SERP API to identify relevant company websites with the desired information. To ensure quality results, the script applied a RAG architecture to parse page content and isolate relevant text. This was then passed to a LLM for semantic analysis. The tool accepted CSV inputs of company domains and returned structured data for outreach, significantly improving the scalability of the growth team’s operations. Key engineering challenges included optimizing runtime for deployment and handling asynchronous requests. This project highlights how automation can reduce manual effort in critical workflows.
The STS research paper investigates the impact of AI-powered hiring tools and how they can unintentionally discriminate against certain groups. The Actor-Network Theory (ANT) framework was used to explore how three main groups - AI developers, employers, and regulators - interact to shape the outcomes of these systems. Studies showed that these hiring tools were often subject to bias that creeps in through historical hiring data and gets baked into algorithms that are supposed to be objective. Real-world examples like Amazon’s AI resume screener, which penalized applicants for mentioning “women,” showed how easily these systems can go wrong when no one steps in to question or audit them. The ANT framework demonstrated how no single actor is solely responsible - bias comes from the relationships and assumptions built into the entire system. This paper argues that developers need to be more transparent, employers need to stay involved instead of deferring blindly to AI, and regulators need to step up with clearer rules.
These two projects offer insights into the benefits and limitations of automation in decision-making systems that influence professional opportunity. While the technical work demonstrated how automation can improve efficiency and scale in talent outreach, the STS research highlighted how similar tools - when deployed without transparency - can have negative effects. Both projects successfully demonstrated the importance of ethical responsibility in building automated systems that affect people’s lives and met the expectations set earlier in the year.
