____ _ _ _ ___ _ __ ____ ___ _ ____ _
/ ___|| | | | | / _ \| |/ / / ___|_ _| | | _ \ / \
\___ \| |_| | | | | | | ' / | | _ | || | | | | |/ _ \
___) | _ | |__| |_| | . \ | |_| || || |___| |_| / ___ \
|____/|_| |_|_____\___/|_|\_\ \____|___|_____|____/_/ \_\
Research Scientist @ Meta · Ph.D. in Computer Science, University of Florida (2026) NLP × Cybersecurity · Open-Source Supply Chain Security · Trustworthy AI Evaluation
At Meta, I own decision quality and observability for an ML agent that verifies advertiser identity at scale. That means building the evaluation harness behind it: precision/recall regression tests against human-labeled ground truth, and production dashboards tracking error rates and business impact.
Before that, my Ph.D. asked a question that sounds soft but turned out to matter: does the way developers talk to each other change how secure their software is? (Short answer: yes.)
+ Research Scientist @ Meta: evaluation & observability for ML verification agents
+ PC member, USENIX Security 2027
+ Reviewer, ACL Rolling Review
+ Under review @ ICSE 2027: two papers on dependency-vulnerability remediation in npm
+ Under review @ ARR: OSS-Stance, a stance-detection benchmark for OSS discussionsDissertation: How Communication Dynamics Shape Vulnerability Management in Open-Source Software. Advised by Dr. Bonnie Dorr.
| 🔍 FORCE | Framework for Open-source Risk and Community Evaluation: correlates how vulnerabilities evolve over time with community signals (sentiment, toxicity, stance, outrage) |
| 🧠 CBPE | Cyber Behavior Pattern Extractor: a neuro-symbolic framework that makes LLMs more factually reliable on Cyber Threat Intelligence and builds temporally-aware knowledge graphs (IARPA-funded) |
| 📦 Supply chain | Measuring whether npm vulnerability fixes actually stick, and bounding false CVE-to-dependency attributions without an oracle |
| 💬 Online harms | Detecting subtle toxicity and studying how people engage with misinformation |
publications ██████████████ 14 peer-reviewed papers
patents ██ 1 US patent + 1 provisional
📚 Full list on Google Scholar and on my website.
* 2026-06 (HEAD -> main) Research Scientist @ Meta
* 2021-01 Graduate Research Asst. @ University of Florida [Ph.D. '26, M.Sc. '22]
|\
| * 2024-01 AI Resident @ SandboxAQ
| * 2023-06 Research Intern @ Accenture Security Labs
|/
* 2018-06 Software Engineer @ Ost → Pepo → Moxie
* 2017-06 Research Intern @ IBM India Software Labs
* 2014 (tag: v0.1) B.E. Computer Engineering @ University of Pune
expand commit messages
- Meta: decision quality, observability, and evaluation for an advertiser-verification ML agent
- SandboxAQ: ML pipeline classifying cryptographic strength across 8 algorithms on 300K+ encrypted files (0.79 AUC)
- Accenture Security Labs: analyzed 100K+ commits across 20 OSS repos to find malicious developer patterns
- Ost → Pepo → Moxie: one founding team, three pivots; Ethereum token platform API at 500+ tx/s, feed ranking (+35% engagement), real-time video and chat
- IBM India Software Labs: Hyperledger Fabric IAM with zero-knowledge authentication; co-authored a US patent on identity attribute trust scoring
drwxr-xr-x languages/ Python JavaScript Node.js C/C++ SQL
drwxr-xr-x ml/ PyTorch TensorFlow scikit-learn spaCy LLM evaluation
drwxr-xr-x data/ MySQL DynamoDB Neo4j Cassandra
drwxr-xr-x research/ NLP Knowledge Graphs Neuro-symbolic AI OSS Security
- 🏆 Best Paper Award, SocialSec 2023
- 🎤 Predicting OSS Vulnerabilities Through Communication Analysis, OpenSSF Community Day NA 2025
- 🎟️ Travel Grant, Linux Foundation Open Source Summit 2025
- 🎓 Student Conferenceship, ACSAC 2022 · Student Travel Grant, IEEE S&P 2021 · Student Grant, USENIX Enigma 2021
> status: shipping evals @ Meta
> location: Menlo Park, CA
> fun_fact: I studied toxic online behavior for my Ph.D., but I promise I'm nice IRL