Ravi Sharma

Computer Vision Researcher · Math AI Tools Reviewer · STEM Educator

6+ Years in AI

50+ Tools reviewed

90%+ Accuracy tested

IIT Delhi CS Eng, B.Tech

“Photo-based math solvers are only as good as their vision layer — and most of them are let down by OCR that can’t handle real-world conditions: handwriting, shadows, curved pages, subscripts. I test every tool on the problems students actually photograph, not just clean screenshots. That’s the only standard that matters.”

— Ravi Sharma, Math Camera Solver

About

Computer vision researcher who tests math AI on real-world input

Ravi Sharma is a computer vision researcher, STEM educator, and AI tools reviewer with six years of experience across optical character recognition, math reasoning models, and educational technology. He holds a B.Tech in Computer Science Engineering from IIT Delhi, where he specialized in image processing and machine learning — the same technical domains that underpin every photo-based math solver he reviews.

Before joining Math Camera Solver, Ravi worked for three years as an applied ML engineer at an EdTech startup, building and benchmarking OCR pipelines specifically for handwritten STEM content. That role gave him a precise technical understanding of where vision-based math solvers fail: handwriting ambiguity, symbol segmentation errors, curved textbook pages, and low-contrast images. He knows exactly what to stress-test, and what the failure modes look like at each step of the pipeline.

At Math Camera Solver, Ravi evaluates photo-based math tools by running them through the kinds of inputs students actually submit — messy handwriting, phone photos with shadows, screenshots of printed problems, and edge-case symbol combinations. His reviews focus on OCR accuracy, step-by-step explanation quality, and whether the tool is genuinely useful for students working from physical textbooks at 11pm before an exam.

Expertise

Photo-Based Math Solvers OCR & Vision AI Testing Handwriting Recognition Step-by-Step Logic Quality Photomath Alternative Reviews Algebra & Calculus Accuracy Computer Vision (ML) Image Processing STEM EdTech Reviews Word Problem Parsing Real-World Input Testing Cross-Device Usability

Education & Background

🎓

B.Tech Computer Science Engineering — IIT Delhi

Specialization in image processing, machine learning, and optical character recognition. Final-year project on handwritten mathematical expression recognition — directly relevant to the OCR challenges in photo-based math solvers.

⚙️

Applied ML Engineer — EdTech Startup (3 years)

Built and benchmarked OCR pipelines for handwritten STEM content. Developed systematic testing protocols for vision model accuracy across different input conditions: lighting, handwriting styles, paper types, and symbol density.

🔬

Independent Math AI Researcher (3 years)

Systematic benchmarking of photo-based math solvers including Photomath, Symbolab, Mathway, Microsoft Math Solver — across algebra, calculus, geometry, and statistics at multiple difficulty levels.

✍️

Staff Writer & Reviewer — Math Camera Solver

Tests photo-based math tools on the kinds of inputs students actually use — handwritten homework, textbook photos, screenshots with shadows, multi-step calculus problems. Evaluates OCR accuracy, step-by-step explanation quality, and real-world usability across phone, tablet, and desktop.

Review methodology

How every photo math tool on this site is tested

Real-world photo input

Problems are photographed from actual textbooks and handwritten notes — under different lighting conditions, angles, and paper types — not just uploaded as clean screenshots.

Handwriting stress test

Multiple handwriting styles are tested — neat, messy, left-handed, non-native writers — specifically targeting ambiguous character pairs: 1/l, 0/O, x/×, z/2, and fraction bars.

OCR accuracy audit

The extracted problem text is compared against the original input — measuring how many symbols, numbers, and variables were read correctly before the solution is even generated.

Solution correctness check

Final answers are verified against hand-calculated solutions across algebra, geometry, calculus, and statistics — at middle school, high school, and university difficulty levels.

Step logic evaluation

Each solution step is reviewed for logical correctness and pedagogical value — does the explanation teach the method, or just label the arithmetic without context?

Device & browser testing

Tools are tested on iOS Safari, Android Chrome, and desktop browsers — because most students use their phone camera, and mobile UX problems are just as important as accuracy.

💡

Why real photos matter: Most math solver benchmarks use clean screenshots. Real students photograph textbooks in poor lighting with shaky hands. Ravi’s test set specifically includes these conditions — because that’s what the solver actually has to handle.