Computer-science student working
two tracks
: cybersecurity and applied AI.
I build agents that talk, models that predict, and tools that defend, mostly
in Python. I learn by shipping, breaking things on purpose, and reading the
stack trace.
Crop-yield forecasting from public agricultural data. Regression models trained on
government datasets predict yield for a region and suggest planting windows, so
farmers and stakeholders can plan crops with data instead of guesswork.
Built with Vihaan S for the CBSE Science Exhibition 2024–25.
Personal-finance analysis that uses machine learning to surface spending patterns,
flag unusual transactions, and suggest budget adjustments, built as a Social
Action Project.
A security toolkit: modular encryption and decryption, password-strength
analysis, JWT inspection, and educational cracking simulations. Designed for
both CLI and web use, with ethical security testing in mind.
In-browser smishing and scam message detector. Runs a fine-tuned DistilBERT ONNX
safetensors model locally in WebAssembly, paired with SMS screenshot OCR and
sender-header verification to expose phishing without exposing private messages.
// shipping for users who can't see the screen, best constraint i've built under.
Kenpath Technologies
software engineering intern · ongoing
End-to-end mobile engineering with a social-impact brief, accessibility-first
Android apps for visually challenged users, built native and cross-platform
with Kotlin, Expo, and React Native.
Agri-tech platform bridging mobile apps and custom microcontroller hardware (ESP32, Arduino Nano), including AI agents that help farmers identify livestock disease.
Voice-first study companion for university curricula: LiveKit real-time conversation and telephony, with document-retrieval tool calling.
Hardened company infrastructure, phishing-email protection, biometric-linked authentication, and strict role-based access control.
Led AG-UI interface work so highly technical backends land as simple, intuitive UIs.
Toolkit
What I'm hands-on with right now.
Python
My daily language, typed and async. Everything from ML pipelines (pandas, scikit-learn) to security scripting and agent backends.
LiveKit
Real-time voice agents: STT → LLM → TTS pipelines with interruption handling, tool calls, and low-latency audio rooms.
Pydantic AI
Typed LLM agents, structured outputs, tool calling, and validation that catches model drift at the boundary.
ag-ui
The protocol layer between agents and interfaces: streaming tool calls, state sync, and generative UI in the browser.
Skills
Security
Network fundamentals
Cryptography & hashing
Password security
Web & API security basics
Linux hardening
AI & ML
Supervised modelling
Feature engineering & evaluation
LLM tool use & structured output
Voice & conversational agents
Prompt design & eval loops
Engineering
Python, typed, async
HTML, CSS, JavaScript
Git & version control
Linux command line
REST APIs & SQL
Certifications
TryHackMe, SEC0
TryHackMe, Pre-Security
TryHackMe, Advent of Cyber 2024
Times of India, Certified Young AI Explorer
More in progress
Highlights
Competitions, olympiads, and the teaching habit.
HPE CodeWars 2025
Rank 36, 99.41st percentile, at Hewlett Packard Enterprise's coding competition.
Schoolhouse.world
Top-10% volunteer tutor, 45+ hours across 68+ learners in 20+ countries, senior tutor in the SAT Bootcamp, 100% Algebra 2 mastery, 120+ positive peer ratings.
Off the clock: Carnatic vocal and western keyboard, three years of formal training
each, music production, and the occasional puzzle hunt (1st place, intra-school
Computer Matrix).
Contact
Always happy to trade notes on security, agents, and ML. Departures below.