I build systems that make decisions — not just interfaces.
Second-year CSE (AI & ML) student at VIT Vellore, developing real-time autonomous vehicle software at Team Ojas Racing (Formula Student EV). My work sits at the intersection of perception pipelines, explainable ML, and low-level C++ — where software has to be correct, not just functional.
Focus Areas
├── Autonomous Systems & Robotics (ROS2 Jazzy, sensor fusion, perception)
├── Machine Learning & Explainable AI (Isolation Forest, SHAP, behavioral modeling)
├── System-Level C++ Engineering (real-time decision engines, optimization)
└── Embedded Systems (Jetson Orin Nano, Arduino, RealSense, SBG IMU)
Autonomous Software Developer & Operations Team Member (CR)
Building the software stack that makes the car think — and the team that makes it run:
| Domain | Responsibility |
|---|---|
| 🧠 Architecture | ROS2 Jazzy modular autonomous system design |
| 👁️ Perception | Sensor fusion — Intel RealSense stereo depth + SBG Ellipse IMU |
| ⚡ Decision | Real-time autonomous planning & control systems |
| 💻 Compute | Nvidia Jetson Orin Nano — edge AI inference & onboard processing |
| 🔒 Reliability | Safety-critical environment engineering |
| 📋 Operations | Team coordination, race logistics & cross-role ops (CR) |
Hardware stack in use:
Intel RealSense D435i ──┐
├──▶ ROS2 Jazzy (perception nodes) ──▶ Decision Stack
SBG Ellipse N IMU ──┘ │
▼
Nvidia Jetson Orin Nano
(onboard edge compute)
🔐 Insider Threat Dashboard — Internal Data Leak Risk Detector Python · Streamlit · ML · XAI
A full-stack ML surveillance system for detecting insider threats before they happen
A production-grade Streamlit dashboard + simulation suite for insider-risk monitoring. Ingests daily user activity (USB interactions, email metadata, psychometric profiles) and flags anomalous behavior with explainable AI insights.
Core ML Pipeline:
- Isolation Forest trained monthly on cumulative user activity features
- SHAP TreeExplainer provides per-user, per-feature risk attribution
- Dynamic + hard-limit policy checks (USB insertions, sensitive file ratios, external email ratios)
- Severity classification:
Critical → High → Elevated → Normalvia threshold quantiles
Data Model:
- Email signals:
total_emails,external_emails,attachments_sent,bcc_in_email,avg_email_size - USB signals:
usb_insertions,files_accessed,sensitive_files_accessed - Big Five psychometrics (
O,C,E,A,N) — withCandAinverted for risk alignment
System Architecture:
Daily CSV ingestion → Cumulative aggregation → Anomaly scoring
↓ ↓ ↓
Simulation engine Monthly retraining SHAP logging
↓ ↓ ↓
Streamlit dashboard ← Flagged users + risk explanations
Advanced Features:
- Synthetic multi-month data generation (
full_generator.py) - CLI orchestration engine for automated month-cycling & retraining
- Web3 integration blueprint — wallet-auth, on-chain threshold governance, IPFS alert anchoring
🤖 Scrape-Krunch — Real-Time News Intelligence Pipeline Python · Ollama · LLM
Multi-domain news scraper feeding a local LLM for contextual analysis
End-to-end pipeline that turns raw web noise into structured intelligence:
- Scrapes multi-domain news sources in real time
- Feeds structured, normalized data into a locally hosted LLM via Ollama
- Delivers contextual summarization, cross-domain correlation, and on-demand analysis
- Fully offline-capable — no external API dependencies
💰 Adaptive Loan Repayment Scheduler C++ · Optimization · Decision Systems
Dynamic priority-based financial decision engine
A C++ system that acts as an automated financial advisor, reasoning over multiple loans simultaneously:
- Multi-factor decision logic: interest rate, remaining tenure, outstanding balance
- Dynamic repayment priority recomputed at each cycle
- Outputs the globally optimal repayment sequence under budget constraints
🤲 Hand Tremor Stabilization Device C++ · Arduino · Embedded · Assistive Tech
Low-cost real-time assistive device for Parkinson's & essential tremor patients
An embedded hardware system that restores independence for people with involuntary hand movements — stabilizing spoons and brushes in real time:
| Component | Role |
|---|---|
| MPU6050 | 6-DOF motion sensing — detects & classifies tremor signatures |
| Arduino Nano | Embedded control loop & signal processing |
| Servo Motor | Real-time physical counterforce stabilization |
| USB Power | Portable & accessible form factor |
Signal flow: Sensor reads → Tremor classification → Servo actuation (continuous loop)
Languages
AI / ML & XAI
Isolation Forest · SHAP · Behavioral Modeling · Streamlit
Autonomous & Embedded
ROS2 Jazzy · Arduino · Nvidia Jetson Orin Nano · Intel RealSense D435i · SBG Ellipse IMU · Ollama · Decision Systems
🥇 Gold Badge Problem Solver CodeChef
📡 Active Member IEEE-VIT
🏁 Autonomous Software Developer Team Ojas Racing — Formula Student EV
📋 Operations Team Member (CR) Team Ojas Racing — Formula Student EV
I build systems that make decisions — not just interfaces.