Saketh Reddy

DARPA AIxCC Finalist with Team Atlanta (https://team-atlanta.github.io).
Hybrid Fuzzing w/ LLMs

Researched LLM-enabled hybrid analysis techniques for automated vulnerability detection.

Conducted literature review of state-of-the-art program analysis techniques for automated vulnerability detection, covering 36 academic papers. Topics included Symbolic Execution, Directed Fuzzing, Hybrid Fuzzing, Concolic Execution

Created LLM-based directed fuzzing mutator using Rust, Python, LibAFL, and Claude 3.5 Sonnet, outperforming sota AFLGo in stb image time-to-reach function… Show more

Researched LLM-enabled hybrid analysis techniques for automated vulnerability detection.

Conducted literature review of state-of-the-art program analysis techniques for automated vulnerability detection, covering 36 academic papers. Topics included Symbolic Execution, Directed Fuzzing, Hybrid Fuzzing, Concolic Execution

Created LLM-based directed fuzzing mutator using Rust, Python, LibAFL, and Claude 3.5 Sonnet, outperforming sota AFLGo in stb image time-to-reach function benchmark by leveraging semantic understanding of application inputs

Developed program analysis LLM agent using langchain and Reason-Action (ReAct) strategy, solving 84% of relevant CSAW CTF Quals challenges from NYU CTF dataset in <3 minutes per challenge; leveraged custom dockerized bash, Python, and radare2-based tools to reason about and interact with source and compiled code. Show less

Rapidly iterated and deployed applications to enhance TIAA's Cybersecurity risk assessment pipeline workflows.

Designed and implemented on-premises chatbot framework leveraging RasaNLU to enable rapid application onboarding and knowledge base querying within TIAA's Cybersecurity application teams. Solution was designed to dynamically query and iteratively fine-tune on information organized within levelled scopes-- allowing individual chatbots to specialize in responding to inquiries… Show more

Rapidly iterated and deployed applications to enhance TIAA's Cybersecurity risk assessment pipeline workflows.

Designed and implemented on-premises chatbot framework leveraging RasaNLU to enable rapid application onboarding and knowledge base querying within TIAA's Cybersecurity application teams. Solution was designed to dynamically query and iteratively fine-tune on information organized within levelled scopes-- allowing individual chatbots to specialize in responding to inquiries within their specific application domains, while enabling all deployed models to share an underlying core knowledge base.

Led a team of 10 interns to win 1st place in the TIAA 2023 Intern Hackathon by ideating, iterating, and developing a novel GenAI-copiloted Social Media marketing platform capable of bridging the generational gap and allow the company to more effectively educate younger generations on retirement topics. Show less

Developed Retroperm (https://github.com/SpiritSeal/retroperm), an experimental binary analysis framework implementing the angr static analysis engine to retroactively generate manifest-like permissioning information for compiled programs based on the system resources they will attempt to access when executed.

Led a team over 12 months to develop "Bubble Map" (www.bubblemap.app), a collaborative AI-powered mind mapping software that leverages Natural Language Processing and GPT-3 to automatically suggest new ideas to the user based on previously entered data. The app also takes advantage of Physics force simulations to automatically reposition nodes in an aesthetically pleasing yet functional manner, removing inhibiting distractions and enabling users to take advantage of the idea-generation… Show more

Led a team over 12 months to develop "Bubble Map" (www.bubblemap.app), a collaborative AI-powered mind mapping software that leverages Natural Language Processing and GPT-3 to automatically suggest new ideas to the user based on previously entered data. The app also takes advantage of Physics force simulations to automatically reposition nodes in an aesthetically pleasing yet functional manner, removing inhibiting distractions and enabling users to take advantage of the idea-generation capabilities of mind mapping to its fullest. This novel positioning system also drastically reduces the amount of costly database writes the app requires, enabling us to support an eternal, unrestricted free-to-use model that no other collaborative mind mapping platform on the market provides.

In addition to my role as a technical developer, I also served as the project lead, leveraging skills gained from Stanford's ENGR145 to manage many critical facets of the project's ideation and development, including Market Research, Minimum Viable Product planning, User Testing, Feedback Iteration, Project Deployment, etc. During my time in this role, I worked closely with my teammates to ensure we planned and reached critical project milestones over the course of this project. Show less

Paid Internship at Discount Tire Company (DTC) as an Internet of Things (IoT) Device Developer

I developed IoT devices using ESP32s and various sensors to automatically track the amount, timestamp, and duration of vehicles that entered Discount Tire’s service bays, and then send that data over the air into DTC’s SAP HANA database. This data could then be used to track the efficiency of the various bays in different stores in order to determine general and specific trends.

During… Show more

Paid Internship at Discount Tire Company (DTC) as an Internet of Things (IoT) Device Developer

I developed IoT devices using ESP32s and various sensors to automatically track the amount, timestamp, and duration of vehicles that entered Discount Tire’s service bays, and then send that data over the air into DTC’s SAP HANA database. This data could then be used to track the efficiency of the various bays in different stores in order to determine general and specific trends.

During my time on this project, I was able to develop my technical skills in Software Engineering, Electronics, 3D modeling, and additive manufacturing (3D printing) for mass production. Additionally, I was also able to improve some of my other skills over this period, including rapid idea generation, prototyping, and communication.

While I was employed at DTC, I also briefly worked on other exciting projects the company was developing, including a digital signage solution for contactless drive-through vehicle servicing and an NLP-driven chatbot to assist customers on their website. Show less

A mock tech startup that I worked in during Stanford's 2021 Technology Entrepreneurship Summer Session Course (ENGR 145S).

As a Technical Lead, I gained experience in a wide variety of tech skills as they were needed. The scope of the tasks I worked on ranged from preparing a website using [HTML/ CSS/ JS], to deploying it onto a Namecheap domain via GitHub Pages, to developing a mockup app for the startup using Flutter and Firebase.

In addition, during my time with the group, I… Show more

A mock tech startup that I worked in during Stanford's 2021 Technology Entrepreneurship Summer Session Course (ENGR 145S).

As a Technical Lead, I gained experience in a wide variety of tech skills as they were needed. The scope of the tasks I worked on ranged from preparing a website using [HTML/ CSS/ JS], to deploying it onto a Namecheap domain via GitHub Pages, to developing a mockup app for the startup using Flutter and Firebase.

In addition, during my time with the group, I developed essential skills regarding communication, team management, and project planning. Show less

Worked with Dr. Sat Parashar [Adjunct Faculty, Rady School of Management, University of California, San Diego] to collect and analyze of data regarding COVID-19 deaths in the USA to better understand the impact of the virus over time.