Computer Science @ ASU • Cybersecurity

Hi, I’m Anish — I build secure, human centered software.

Interested in AI safety, embedded ML, and ethical hacking. I like turning research ideas into usable tools and sharing what I learn along the way.

View projects Get in touch

Selected Projects

Click a project to expand its write-up.

> Nebulo

Collaborative student platform combining security, automation, and a touch of humor where ideas actually turn into projects.

JavaScript Next.js Prisma Socket.IO Python - Flask PostgreSQL

Click to expand →

🌐 Non-Technical

Imagine a place where university chaos meets organized brilliance, that’s Nebulo. It’s like if Discord, GitHub, and your group chat had a well behaved child who actually finished projects. Students can post ideas, find teammates, apply to projects, and instantly get dropped into secure chats where brainstorming doesn’t die after the first “hey.”

Whether you’re designing a robot, building an app, or just trying to find someone who knows what a “backend” is: Nebulo turns the awkward “we should totally collab sometime” into “wow, we actually built this thing.” It’s your creative playground, startup incubator, and caffeine substitute all rolled into one.

It’s not about who’s smartest; it’s about who’s curious enough to start. Nebulo makes sure every idea from “let’s detect sirens with ML” to “what if the whiteboard could talk?” finds the right people to make it happen.

💻 Technical

Nebulo is a distributed collaboration platform engineered for scalability, security, and seamless team formation. Built with a modular architecture, it supports project creation, application workflows, threaded discussions, and automatic group chat generation within isolated project instances.

Each chat instance leverages Diffie-Hellman key exchange (DHKE) for secure peer-to-peer communication, with encrypted message persistence handled through a hybrid client server model. The authentication system employs salted hashing and tokenized sessions to prevent credential replay, while project data is normalized through a custom schema optimized for asynchronous access and cross user querying.

On the backend, Nebulo’s API layer mediates between a real-time messaging service and a RESTful microservice core, allowing for clean separation between transactional data and live interaction. Future iterations include implementing federated learning models to provide project recommendations without centralized user profiling, role-based access control (RBAC) for granular permission management, and an ML driven moderation layer capable of detecting behavioral anomalies and flagging potential toxicity patterns.

In essence, Nebulo combines the transparency of open collaboration with the rigor of secure engineering: a platform built not just for students, but for the future of how teams form, build, and innovate together.

View on GitHub

> Ember

A 400M parameter character level Transformer built for deep linguistic understanding, model interpretability, and efficient large scale training.

Python PyTorch Transformer Neural Net Multi-GPU

Click to expand →

Ember is a 400 million parameter character level Transformer language model engineered to push the boundaries of compact intelligence and interpretability. Unlike traditional word level systems, Ember learns language from its atomic structure, letter by letter, uncovering the statistical and semantic patterns that emerge long before syntax ever does.

Built on a custom multi-GPU training pipeline, Ember employs adaptive layer scaling, mixed precision optimization, and context aware token scheduling to achieve high throughput without sacrificing numerical stability. Every component, from tokenizer to optimizer was tuned for clarity and control, turning Ember into both a performant model and an open research instrument.

Its supporting framework includes granular perplexity analysis, activation tracing, attention head visualization, and automated ablation workflows that map out how meaning, grammar, and structure evolve across depth. Ember also supports quantized and distilled inference for deployment testing, bridging large scale modeling and edge compatible experimentation.

In essence, Ember is a study in computational introspection, a system that not only generates text, but reveals how machines learn to write, predict, and imagine from scratch. It stands as one of the most technically ambitious and research oriented projects in the Nebulo ecosystem.

View on GitHub

> Siren Detector

Embedded DSCNN based ML system for emergency siren detection: 4M param teacher distilled into a 38K param C++ micro model running on ESP32 hardware.

C++ ESP32 DSCNN TFLite Micro Knowledge Distillation

Click to expand →

Siren Detector is an embedded machine learning pipeline built around a Depthwise Separable Convolutional Neural Network (DSCNN) architecture, designed to achieve high auditory precision with minimal computational cost. The system employs a two stage distillation process: a 4 million parameter teacher network, trained in TensorFlow, generates refined supervision data used to train a compact 38K parameter student model implemented in pure C++ with TensorFlow Lite Micro.

The result is a highly efficient neural architecture capable of distinguishing emergency sirens from complex ambient soundscapes in real time, all on an ESP32 microcontroller running at just a few hundred kilobytes of memory. Despite its size, the distilled model preserves much of the perceptual acuity of its larger counterpart, thanks to knowledge transfer and aggressive quantization strategies.

Siren Detector now operates on a custom hardware board, performing continuous low latency inference to support a larger safety and automation framework. The project demonstrates how deep learning can escape the cloud: compressing intelligence from millions of parameters into something small enough to run anywhere, yet smart enough to matter.

View on GitHub

Experience

Research, teaching, and leadership across academia and cybersecurity.

Undergraduate Researcher — PERSUE Labs

Arizona State University · Jan 2025 – Present

Conducting research on data security in Android and Zoom ecosystems, identifying privacy leak vectors and exploring mitigation techniques for mobile platforms. Collaborating with faculty and peers to design secure communication tools and publish reproducible results.

Undergraduate Teaching Assistant — FSE 150

Arizona State University · Aug 2025 – Present

Mentor 20+ students weekly on problem-solving and teamwork. Assist professors in designing course materials and active learning tasks for “Grand Challenges for Engineering.”

Program Coordinator — Digisafe

Bangalore, India · Apr 2023 – May 2023

Led corporate cybersecurity awareness training, performing live hacking demonstrations and producing professional video tutorials integrated into programs reaching 500+ employees.

President — High School Coding Club

Bangalore, India · 2022 – 2023

Organized hackathons, CTFs, and weekly algorithm challenges for 30+ members. Spearheaded inter-school computer science quiz events and fostered peer learning in programming and cybersecurity.

Cybersecurity

Active practice, competitive CTFs, and applied research in offensive & defensive security.

> Shellphish — CTF Team

Team member — 10+ national & international CTFs

Compete on Shellphish in Jeopardy and Attack-Defense style events. Responsibilities include reverse engineering, pwn challenges, exploit development, and post-contest writeups. Regularly collaborate on team retros and share tooling that speeds up triage and patch validation.

Focus: reverse engineering · pwn · KOTH · automation

> Applied Security & Certifications

Ethical Hacking certification · Practical pentesting experience

Completed formal ethical hacking training and applied VAPT techniques across research projects and corporate awareness programs. Familiar with threat modeling, responsible disclosure, and building reproducible PoCs for vulnerabilities.

> Outreach & Impact

Community teaching · Awareness campaigns

Delivered live demos and corporate training sessions (Digisafe), authored internal awareness materials. Passionate about making security approachable and reproducible for learners.

> ASU Hacking Club

Member & active contributor

Contact

Open to internships, research, and collaborations.