krissh = {
    "name": "Krissh Verma",
    "education": "B.Tech CSE (AI & ML) @ VIT Vellore",
    "batch": "2023–2027",
    "academic_track": {
        "cgpa": 9.06,
        "rank": "Top 7 / 300"
    },
    "primary_direction": "Software Engineering",
    "specialization": "Applied AI & Machine Learning",
    "engineering_interests": [
        "backend systems",
        "computer vision",
        "machine learning pipelines",
        "full-stack products",
        "developer tooling"
    ],
    "outside_code": ["gaming", "video editing", "fitness"],
    "current_goal": "Build systems that are technically strong, usable and production-minded"
}

I am a software engineering student at VIT Vellore, specialising in AI and Machine Learning. I enjoy working where software, systems and intelligent models meet—whether that means designing a gesture-driven desktop interface, building an anomaly-detection pipeline or shipping a platform used by thousands of participants.

My approach is simple: understand the constraint, build the system, test it under realistic conditions and improve the parts that break.

Python · MediaPipe · OpenCV · PyAutoGUI · pycaw · pystray · psutil

Explore this project on my portfolio

A webcam-based human-computer interaction system designed to replace conventional mouse input through real-time hand-landmark tracking.

• Supports cursor movement, click and drag, scrolling, zoom and screenshots.
• Uses a dual-hand control protocol instead of overloading one hand with every action.
• Maps left-hand mode selection to right-hand motion for volume, brightness, application switching and zoom.
• Integrates with operating-system audio, tray and power-aware services.

Engineering focus: designing gesture states that remain usable in real time while preventing accidental commands and conflicting actions.

Python · Scikit-learn · SVM · Autoencoders · Mahalanobis Distance

Explore this project on my portfolio

An ML-assisted diagnostic pipeline for detecting airflow leaks in diesel engines operating under stable load.

• Uses a synthetic, physics-guided dataset to model airflow anomalies across engine subsystems.
• Combines Mahalanobis distance, a one-class SVM and four autoencoder architectures.
• Produces a go/no-go diagnostic result, confidence score and probable leak location.
• Treats the task as both anomaly detection and subsystem isolation rather than simple classification.

Engineering focus: combining multiple imperfect detectors into a practical diagnostic decision instead of relying on one model score.

Next.js · TypeScript · MongoDB · Tailwind CSS · Vercel

Explore this project on my portfolio

A team-built event platform for VIT’s flagship entrepreneurship fest, designed for high-concurrency registrations and live participation.

• Supported more than 2,000 registrations with zero reported downtime.
• Independently engineered a real-time quiz experience used by more than 1,000 participants.
• Added live leaderboards and scalable MongoDB-backed result storage.
• Used server-side rendering and Vercel deployment for responsive delivery during the event.

Engineering focus: keeping scoring, leaderboard state and participant access reliable under concurrent traffic.

Software Engineering
├── Reliable backend services and APIs
├── Full-stack products with measurable users
├── System design and production deployment
└── Strong data structures and problem-solving foundations

Applied AI / ML
├── Computer vision and human-computer interaction
├── Anomaly detection and diagnostic systems
├── Model evaluation beyond headline accuracy
└── AI features integrated into usable software

• Top 7 out of approximately 300 students in CSE (AI & ML) at VIT.
• Senior Committee Technical Member at Entrepreneurship Cell, VIT Vellore.
• Helped build and operate technical infrastructure for an event platform serving thousands of students.
• Organised logistics and scheduling across 10+ events with more than 500 attendees.
• Completed coursework in DevOps, Agile, Design Thinking and supervised machine learning.

When I am not coding, I am usually gaming, learning video-editing techniques or working on fitness and content ideas. Gaming keeps me interested in interactive systems; editing has made me more attentive to pacing, presentation and the experience around a product—not only the code behind it.