I build tools thathold up underadversarial pressure.
Master's in Cybersecurity at UNC Charlotte, teaching assistant for ITIS 6200, and a member of the FORTRESS lab. Before that: four years of computer science at VIT Bhopal, an AI internship at DRDO, and three peer-reviewed papers.
Approach
Security is a property of systems, not a layer you add to them.
I care about the layer where a design decision becomes an exploitable one. That has meant writing a password manager in C with no dependencies, encrypting ELF binaries so they only ever decrypt in memory, and building a static analyser that reads C/C++ the way an attacker reads it.
Lately the same instinct has pointed inward, at models rather than binaries. At UNC Charlotte I pitched ARGUS — adaptive safety for LLMs through mechanistic interpretability — and trained sparse autoencoders to pull interpretable features out of a model's internal activations. Reverse engineering, with different substrate.
The other half of the work is teaching it. I helped restructure the syllabus for ITIS 6200, wrote the weekly labs, and built deliberately vulnerable environments so students could break something real before being asked to defend it.
Applied cryptography
Authenticated encryption, memory-hard key derivation, and entropy sourcing that does not trust the operating system's RNG. Built into Hyperion and ElfEnc.
Binary & systems security
ELF internals, in-memory execution, Linux kernel surfaces, and the parts of a program that only exist after the compiler is finished with it.
Mechanistic interpretability
Sparse autoencoders over model activations, looking for features that correspond to something a human can name — and then using them for safety.
Static analysis
Lexical tokenisation and signature-driven detection of vulnerability classes in C/C++ source, with patching workflows attached.
Network & protocol analysis
Packet deconstruction down to raw headers across every protocol on the wire, and ML pipelines built on top of the captures.
Digital forensics
Cross-platform log collection with verifiable integrity — modular, lightweight, and built to survive being questioned in review.
Experience
7 positionsUniversity of North Carolina at Charlotte
Charlotte, NCTeaching Assistant
- Helped revise and restructure the syllabus for ITIS 6200 — Principles of Information Security and Privacy.
- Wrote scripts, apps, tools, and weekly labs covering security principles, encryption algorithms, and block cipher modes of operation.
- Built intentionally vulnerable environments for students to experiment against and analyse.
- Ran regular TA sessions walking through exploit classes including XSS and SQL injection.
University of North Carolina at Charlotte
Charlotte, NCResearch Assistant
- Pitched ARGUS — adaptive safety for large language models via internal mechanistic interpretability methods.
- Programmed and trained sparse autoencoders to surface and interpret meaningful internal signals and features.
The Globalizers
Indore, IndiaTechnical Intern
- Developed the company's iOS and Android applications while maintaining security protocols throughout.
- Created machine learning algorithms to maximise student learning rate through personalised data-feeding.
Fllink
Bhopal, IndiaSecurity Analyst Intern
- Monitored and analysed security incidents to identify threats and recommend responses — cutting breach response time by 30%.
- Communicated risk to stakeholders at every level to build a culture of security consciousness.
Defence Research and Development Organisation
Ministry of Defence, IndiaArtificial Intelligence Intern
- Developed neural networks and simulation systems for military intelligence, reaching 92% accuracy on identification assessment.
- Collaborated with a team of scientists on an intelligence-gathering system integrating data from multiple sources.
- Conducted safety research on the project using simulation systems.
- Presented findings and recommendations to senior military officials.
Leadership
Member
- Discussion leader and member of Dr. Xiang's Formal Techniques for Resilient and Secure Systems lab.
- Contribute to the group's ongoing research projects.
- Hold weekly discussions presenting new research gaps and potential implementations.
CyberWarriors
VIT BhopalPresident & Co-founder
- Founded the on-campus cybersecurity enthusiasts club.
- Organised events and educational bootcamps for juniors to learn from upperclassmen.
- Managed VMs, recorded results, and propagated findings across the group.
Selected work
7 projectsHyperion
Post-quantum encryption, memory-safe retrieval, password managementA cross-platform, zero-dependency password manager written in C. XChaCha20-Poly1305 authenticated encryption, a custom 64 MB memory-hard KDF to blunt GPU brute-force, and a hardware-agnostic entropy harvester that seeds a custom CSPRNG from nanosecond-precision keystroke timing — so it never has to trust a compromised OS-level RNG.
- XChaCha20-Poly1305
- Memory-hard KDF
- CSPRNG
- Zero-dependency
Marvin
Log parsing, cross-platform auditing, forensic collectionA lightweight, cross-platform forensic log collection tool built for simplicity, modularity, and verifiable integrity — the properties general-purpose log shippers tend to trade away.
- Forensics
- Log integrity
- Cross-platform
- Modular
ElfEnc
Linux binaries, encryption, cryptographyEncrypts ELF executables behind a password using custom cryptographic primitives — a modified RC4 stream cipher and bespoke key derivation — then executes them entirely in memory, so the static binary on disk never reveals what it runs.
- ELF
- In-memory execution
- Stream cipher
- Key derivation
Peepapp
Decompiling, Android security, privacyA full-scale Python library that decompiles Android applications and uses code signatures to determine exactly which invasive permissions and device access each app actually exercises.
- Android
- Decompilation
- Code signatures
- Privacy
Meowlinter
SAST, lexical tokenization, secure programmingA static analysis tool that identifies potential security vulnerabilities in C/C++ source, and lets the user run code patching and vulnerability assessment simulations against its findings.
- SAST
- Tokenization
- C/C++
- Patching
Attack-ml
Classification, machine learning, packet deconstructionA machine learning pipeline for network data — cleaning, balancing, and analysis — with active sniffing across every protocol in play for comprehensive monitoring.
- ML pipeline
- Classification
- Packet capture
- Networking
GoodSniffer
Networking, WinAPI, packet deconstructionA packet sniffer built to go further than Wireshark on header coverage — capturing every header type and payload in hex, raw, and ASCII simultaneously.
- WinAPI
- Packet analysis
- Hex/ASCII
- Protocols
Publications
3 peer-reviewedCredentials
Education
University of North Carolina at Charlotte
Master's in Cybersecurity · Charlotte, USAVellore Institute of Technology
B.Tech, Computer Science Engineering · Bhopal, IndiaCertifications
Languages
- C
- C++
- Python
- Rust
- Go
- JavaScript
- SQL
- HTML/CSS
Tools & platforms
- Git
- Google Cloud Platform
- AWS
- Azure
- Embedded systems
- IDEs
Technologies
- System programming
- Linux kernel
- Bash scripting
- OSINT
- AI-driven development
Coursework
- Artificial Intelligence
- Information Security
- Secure Systems
- Penetration Testing
- Networking
- Computer Architecture
- System Programming
- Digital Forensics
- Secure Coding Practices
- Reverse Engineering
- Logic and System Design
- Software Development
- Probabilistic Mathematics
Off-hours
The things that keep the pattern-matching sharp.
Music
The Beatles, Pink Floyd and Led Zeppelin at one end; Ninajirachi, Bladee, Ecco2k, Yung Lean and Mechatok at the other. Sixty years apart and both ends are doing the same thing — building something that shouldn't hold together and making it hold together anyway.
Pokémon
Probably the first system I ever took apart to see how it worked. Type charts, damage formulas, IVs and EVs, breeding maths — a game that hands you a rulebook and quietly dares you to find where it bends. I never really stopped, and I'm fairly sure it's why I ended up doing this.
Cars
Men and their musings with the machine. I love the sound of a good rev — a whole engine's worth of geometry and timing resolving into one note you can hear from a street away.
Making things
Mostly audio: producing, arranging, chasing a mix until it sits right. Same loop as everything else here — build, listen for what's wrong, take it apart, go again. Tools, papers, labs, tracks. It's all the same instinct pointed at different material.
Contact
Open to research collaboration and security work.