Building Solutions Through Code & Curiosity
MCA Student · Developer · Analyst
Not just a developer.
I treat engineering like applied science. Every project starts with a problem, not a feature list.
I approach engineering problems like a researcher: define the question, form a hypothesis, run the experiment, measure results.
My current focus is on systems that sit at the intersection of data, language, and product — tools that make information useful, not just accessible.
When I'm not building, I'm breaking things to understand how they work.
Systems thinker
I map dependencies before writing a single line. Complexity is managed, not feared.
Data-driven
Gut feelings get a hypothesis. Hypotheses get experiments. Experiments get measured.
Finish-line focused
Features ship when they solve the problem they were built for — not before, not after.
Curiosity-led
The best projects started as "I wonder if..." Most of mine still do.
Tools that I used / learning.
Projects
Each project is an experiment. Here's what I tried, what worked, and what I'd change.
Face Gesture Controller
Hands-free reels control using real-time gesture and facial expression detection
FamSpace – Offline-First Family Digital Vault
Secure, hierarchical file manager for organizing family data
Expense Insight Engine
Personal finance tracker with ML-powered anomaly detection
Portfolio Website
Showcasing my work and skills in a modern, responsive design
Deepfake Image Detection
Detecting AI-generated faces using machine learning for trustworthy digital media
NDVI–LST Multi-Temporal Analysis (Manipal)
Assessing vegetation–temperature dynamics in a monsoon-driven coastal environment
South Lhonak Glacier Health Assessment
Analyzing glacier velocity and health post-Teesta flood hazard in Sikkim
Moreover: Trying not to be the best, but be someone to Contribute, Try again.
Deep dives.
Selected experiments with a full breakdown — problem to post-mortem.
Face Gesture Controller
Hands-free reels control using real-time gesture and facial expression detection
Controlling Instagram Reels requires constant hand interaction. Wanted a completely hands-free way to like, skip, and navigate short-form content using only facial expressions.
MediaPipe FaceMesh landmarks could reliably detect smile, eyebrow raises, and mouth open gestures with enough stability filtering to avoid false positives — and map them directly to browser keyboard shortcuts.
Built a Python script using MediaPipe FaceLandmarker for 468-point face mesh detection and OpenCV for real-time webcam capture. Gesture stability is enforced by requiring 8+ consecutive frames before firing. PyAutoGUI sends keyboard shortcuts directly to the focused Chrome window — no mouse movement involved primarily focused on Instagram Reels (Arrow keys, L).
Achieved stable real-time gesture detection at 15+ FPS. Three gestures reliably detected — left head tilt for next video, right head tilt for previous video, mouth open for like. Zero mouse interaction needed once Chrome is focused.
Add wink detection for play/pause. Explore eye-tracking for cursor control. Package as a background system tray app so no terminal window is needed.
How I got here.
Not a resume. A log of what I learned — in order.
MCA — Master of Computer Applications
Manipal Institute of Technology, Karnataka, India.
Asian Development Research Institute (ADRI) — Research Intern
Worked on the project "Building upon IHIP-IDSP strengthening surveillance system in Bihar" under department of health, government of Bihar..
Part-time Development Role at SH1ELD Tech
Contributed in Sikkim's local tourism by developing personalized web-applications.
Joined SH1ELD Tech Infosec Solutions (Internship)
Got my first internship and participated in real-world projects.
Bachalor's of Computer Applications
Sikkim Manipal Institute of Technology, Sikkim, India.
Higher Secondary (Class 12)
MC William Higher Secondary School, West Bengal, India.
Things I haven't built yet.
Ideas in progress, shelved experiments, and future directions.
Code Review Copilot
A GitHub bot that summarizes PRs, flags complexity hotspots, and links to relevant docs.
Code review is a high-value but time-consuming ritual. Automation should handle the boring checks so humans focus on architecture.
Live Interview Prep Engine
Adaptive DSA problem recommender based on your weak pattern history.
Spaced repetition is well-studied but rarely applied to coding interviews. Pairing forgetting curves with difficulty ramps should accelerate mastery.
API Contract Drift Detector
CI tool that diffs OpenAPI specs between releases and flags consumer-breaking changes.
API versioning failures are expensive and quiet. Most teams find out when prod breaks, not before.
Self-Hosted Reading Tracker
Notion-free book notes with highlight extraction from Kindle and Kobo.
Kindle highlights rot in the cloud. The knowledge you capture should live where you can search it without a subscription.
Let's work on something.
Open to interesting problems — freelance, collaboration, or full-time roles.
Built with Next.js · Tailwind CSS · Framer Motion
Subhadeep Dhar · 2026