Download PDF | Back to site
JOYSTON MENEZES
Research Interests
Health communication and health misinformation; exposure to harmful content on algorithmically curated platforms; digital mental health; measurement validity in computational social science.
Education
| University of Southern California | Jan 2023 – Dec 2024 |
| Master of Science in Computer Science, Viterbi School of Engineering | Los Angeles, CA |
- GPA: 3.71/4.00. Degree conferred December 18, 2024.
- Coursework: Applied Natural Language Processing; Information Retrieval and Web Search Engines; Database Systems; Analysis of Algorithms; Web Technologies; Multimedia Systems Design.
- Honor: Order of the Laurel and the Palm (2025), USC's highest honor for graduating students, recognizing leadership, scholarship, and service.
| NMAM Institute of Technology | Aug 2017 – Aug 2021 |
| Bachelor of Engineering in Computer Science | Nitte, India |
- Thesis: Detection and Visualization of Bone Abnormalities on X-ray Images Using Convolutional Neural Networks (Best Project of the University, 2021).
- Honors: Best Outgoing Student of the College and of the Computer Science Department; Outstanding Achievement Award (2021).
- Leadership: Student President; Branch Captain representing the Computer Science Department in technical and non-technical events.
Publications
Conference Papers and Proceedings
- Pinto, G., Bickham, C., Salkar, T., Menezes, J., Luceri, L., & Ferrara, E. (2025). Tracking the 2024 US Presidential Election Chatter on TikTok: A Public Multimodal Dataset. In Companion Proceedings of the ACM Web Conference 2025 (WWW '25 Companion), 773–776. [DOI] [arXiv] [Dataset]
- Shetty, S., Shetty, S., & Menezes, J. (2025). Augmenting Large Language Models with Dense Retrieval Techniques for Enhanced QA Performance. In 2025 IEEE International Conference on Distributed Computing, VLSI, Electrical Circuits and Robotics (DISCOVER).
- Shetty, S., Hegde, S., Shetty, S., Shetty, D., Sowmya, M. R., Miranda, R., Sequeira, F., & Menezes, J. (2021). Deep Learning Photograph Caption Generator. In D. P. Shetty & S. Shetty (Eds.), Recent Advances in Artificial Intelligence and Data Engineering: Select Proceedings of AIDE 2020 (Advances in Intelligent Systems and Computing, Vol. 1386). Springer. [Proceedings]
Book Chapter
- Rodrigues, A. P., Menezes, J., Fernandes, R., Aishwarya, Chiplunkar, N. N., & Padmanabha, V. (2021). Identification and Classification of Paddy Crop Diseases Using Big Data Machine Learning Techniques. In A. K. Tyagi (Ed.), Data Science and Data Analytics: Opportunities and Challenges (pp. 47–64). Chapman and Hall/CRC. [DOI]
Acknowledged Contributions
- Pessianzadeh, A., Jamie, P., Sultana, N., Himmelstein, G., Zektser, Y., Ganz, P. A., Hosseinmardi, H., Ghasemian, A., & Rezapour, R. (2026). Beyond Binary Detection: A Multi-Dimensional Taxonomy of Cancer Misinformation on Reddit. In Findings of the Association for Computational Linguistics: EMNLP 2026, 40392–40419. Acknowledged for assistance and contributions to the work. [ACL Anthology] [arXiv]
Research Experience
| UC Davis Department of Communication and University of Warsaw | Aug 2026 – Present |
| Research Collaborator | Remote |
PI: Prof. Magdalena Wojcieszak. Collaborators: Erin Wertz and Subhayan Mukerjee.
- Assessed the feasibility of cross-platform sockpuppet audits of recommendation, search, and AI search features across nine platforms (YouTube, X, TikTok, Instagram, Threads, Bluesky, Truth Social, Gab, GETTR), informing the team's audit planning.
- Built a celebrity-mention matching pipeline over 3.32 million political news posts and 29,999,224 Facebook posts, resolving canonical names from LLM-generated annotations and auditing ambiguous single-token names; matched 1,112,649 posts, reducing the team's political-classifier workload 27-fold.
- Built an outlet typology and per-celebrity concentration measures for a study of how celebrities' political engagement relates to news engagement, and identified 94,349 duplicate records in the source corpus.
- Fit a BERTopic model on a GPU cluster and validated it with a word-intrusion test deployed as a web-based coding task: three human coders identified the intruder word 93.5% of the time on average, with 86 to 92% inter-coder agreement.
| OASIS Lab, Department of Communication, UCLA | Feb 2026 – Present |
| Research Assistant | Los Angeles, CA |
Advisors: Prof. Homa Hosseinmardi and Prof. Amir Ghasemian
- Built an LLM classification pipeline (Python, OpenAI GPT-4o-mini) for a computational study of health misinformation on TikTok, using zero-shot structured outputs to enforce a six-dimension annotation taxonomy, and analyzed the distribution of high-risk medical claims across taxonomy dimensions.
- Contributed to a multi-dimensional taxonomy study of cancer misinformation on Reddit (Findings of EMNLP 2026), acknowledged by the authors for assistance and contributions.
- Implemented eight evaluation metrics comparing Gemini-generated replies with Reddit community replies to cancer misinformation (12,663 responses across 380 posts), including EPITOME empathy classifiers, readability, and an open-source LLM judge (Qwen 2.5) scoring harm and actionability.
- Audited metric validity: showed that a single empathy score hides a split (the model expresses warmth but rarely communicates understanding), rejected a toxicity classifier that flagged correct medical warnings as unsafe, and dropped judge scales that carried no information.
- Built a complete census of children's YouTube channels with more than one million subscribers: 103,066 videos, verified against each channel's reported upload count, with full metadata, a 17,517-video stratified sample, and transcript collection.
| HUMANS Lab, USC Information Sciences Institute | Aug 2024 – Jan 2026 |
| Research Assistant | Marina del Rey, CA |
Supervisors: Prof. Luca Luceri and Prof. Emilio Ferrara
- Built scraping and preprocessing pipelines (Python, Pandas, Apache Spark) covering more than one million TikTok and X posts, plus ten million comments, on the 2024 U.S. presidential election, contributing to a public dataset of 3.14 million TikTok videos (WWW '25 Companion).
- Applied BERTopic topic modeling and VADER sentiment analysis to consolidate more than 700 fragmented topics into 10 coherent themes, and analyzed engagement and sentiment patterns across them.
| Lerman Lab, USC Information Sciences Institute | Aug 2024 – May 2025 |
| Research Assistant | Marina del Rey, CA |
Supervisor: Prof. Kristina Lerman
- Co-developed a hierarchical taxonomy distinguishing harmful from recovery-oriented content on muscle dysmorphia ("Bigorexia") and helped build LLaMA- and Gemini-based classifiers to apply it to TikTok video data at scale, reaching 92% precision.
- Led collection and preprocessing of millions of TikTok videos and cross-platform data, and applied sentiment-aware topic modeling and multimodal embeddings to characterize how harm and recovery narratives co-occur.
| NMAM Institute of Technology | Jun 2020 – Jul 2021 |
| Research Assistant | Nitte, India |
- Developed machine learning models (Decision Tree, Random Forest, SVM, Naive Bayes) to classify six common paddy crop diseases from leaf images; the work was published as a CRC Press book chapter.
- Implemented image preprocessing, Fuzzy C-Means segmentation on Apache Spark, and feature extraction, and benchmarked the speed-up and scale-up of Spark-based against Hadoop-based clustering.
Teaching
| University of Southern California | Aug 2023 – May 2024 |
| Graduate Course Producer, CSCI 585 Database Systems | Los Angeles, CA |
- Supported more than 600 students over two consecutive semesters under Prof. Saty Raghavachary.
- Held office hours, prepared course materials, and helped students work through assignments on data modeling, SQL, NoSQL, and MapReduce.
Professional Experience
| Center for Homelessness, Housing, and Health Equity Research, USC | Jan 2023 – Present |
| Software Developer | Los Angeles, CA |
- Developed five Android applications for field data collection, including the DemoGraphic Count app supporting the Greater Los Angeles Homeless Count.
- Built, hosted, and maintained websites for the Center's projects, including an Olympics homelessness site, Housing LA, and the Homeless Count site, averaging 12,000 to 15,000 visitors per week.
- Produced GIS and spatial visualizations of homelessness patterns (ArcGIS, GeoJSON) and automated data ingestion with Python scripts.
| Alphonso Consultancy | Aug 2021 – Nov 2022 |
| Website Developer and Maintenance Engineer | Udupi, India |
- Built responsive interfaces (Angular, JavaScript) with SQL and PHP back ends, improving operational efficiency by 40%.
| Quality Theorem | Jan 2020 – Aug 2020 |
| Project Lead and Intern | Bangalore, India |
- Led development of an online examination platform (Node.js, MongoDB, REST APIs, Docker).
Patent Applications
Indian Patent Office, published applications; co-inventor. [Nitte patent register] [Patent Office Journal]
- Automated Brain Tumour Diagnosis System and Method Thereof. No. 202441089819 (2024).
- Method for Analysing Public Sentiment on YouTube Comments. No. 202441065225 (2024).
- An Automated Extraction and Summarization System. No. 202441065150 (2024).
- An E-Commerce Product Comparison System. No. 202441065313 (2024).
Honors and Awards
- National Overseas Scholarship, Government of Karnataka, 2024
- Top Data Collector and Surveyor, USC Suzanne Dworak-Peck School of Social Work, 2023
- Student exchange program representative for the Computer Science Department, Ritsumeikan University, Japan, 2020
- Representative of India, Asian Conference of Young Catholic Students, Indonesia, 2019
Service and Community Engagement
| Greater Los Angeles Homeless Count | 2023 – Present |
| Team Lead | Los Angeles, CA |
- Led the DemoGraphic Count initiative, coordinating data collection across Los Angeles to support regional homelessness policy and planning.
| Housing and Water Access Projects | 2024 |
| Volunteer | Mexico and Colombia |
- Tijuana, Mexico (Nov 2024): built homes with Esperanza International. Colombia (Mar 2024): helped build the first water tank for an Indigenous Wayuu community.
| National Service Scheme, Better India Mission | Jul 2017 – Dec 2021 |
| Volunteer Leader | Karnataka, India |
- Implemented household waste management in villages; started the Piggy Bank project supporting homeless individuals and people living with HIV.
Skills
Programming: Python, JavaScript and Node.js, TypeScript, Java, C++, SQL, NoSQL
Methods: NLP and topic modeling (BERTopic), LLM-based classification (LLaMA, Gemini, GPT) with structured outputs, sentiment analysis, large-scale data collection and entity matching, social media data analysis
Tools: Pandas, PyTorch, TensorFlow, scikit-learn, DuckDB, Apache Spark, Git, Docker, ArcGIS
Languages: English, Hindi, Kannada, Konkani, Tulu