In 2024 I chaperoned field trips two days in a row, for two different grade levels.
When I got back to my classroom, roughly 450 ungraded assignments were waiting for me.
After 22 years of teaching I knew the standard move: mark the pile credit or no credit and move on. The students did the practice, they get the points. But any student who practiced it wrong keeps practicing it wrong, and nobody tells them.
I could grind through the pile at night and over the weekend, or let the feedback go. I did not like either option.
That pile is the reason ClassLens exists.
I tried every AI grading tool I could find
I love finding new ways to work efficiently. So my first move was not "stay up later." It was "find the tool that solves this." I tried what was out there. Every product I tried did the same two things.
They graded essays well. Not much else.
I uploaded a screenshot of a student's CAD drawing. The tool sent back a polite paragraph about how it could not interpret the image. I tried a hand-drawn sketch next. Same answer, worded differently. Then a scan of a student's graph, from an assignment that asked the student to reference it. The tool missed the point completely.
The whole category was built for English teachers grading five-paragraph essays. Useful work, real work, but a sliver of what teaching actually looks like for me and my classes.
What teaching actually looks like
I teach engineering. My students turn in:
- CAD drawings exported as PNGs
- Hand-drawn orthographic sketches photographed and uploaded
- Lab writeups that mix paragraphs with data tables and labeled diagrams
- Problem sets with hand-drawn free-body diagrams
- Photographs or screenshots of code with console output
My students also write. Technical reports, résumés, project documentation. The essay-graders worked fine for those. Text in, score out. What broke them was everything else on my pile.
All of it is student work that needs a grade and real feedback by Monday morning.
An English teacher I talked to had the same problem with student notebooks. His kids would turn in ten-plus pages of notes from a unit, and he ended up giving credit-or-no-credit grades because he did not have time to read through them. A science teacher in my building wants to use ClassLens for her own pile. A Spanish teacher I talked to wants an AI that can listen to audio of her students speaking Spanish and check whether the pronunciation and grammar are right. Every department had a stack of student work that did not look like an essay, and the AI grading category had nothing for any of us.
So I built a grading assistant for everyone
I knew what I needed because I needed it for myself. I also knew what other teachers needed because I am the campus Ed Tech. That role puts me in conversation with every department about the tools they use and the ones they wish worked better.
ClassLens grades the file types teachers actually assign. Documents, slides, spreadsheets, PDFs, and photos or scans of student work. Specifically:
- Typed essays and short answers in Google Docs
- Handwritten work, photographed or scanned
- Lab writeups with mixed text, tables, and figures
- CAD drawings and design files exported as images
- Student-drawn graphs, sketches, and diagrams
- Photographs or screenshots of code with console output
- Multi-file submissions where a single assignment includes several of the above
Grading runs on Google Cloud Vertex AI under the Cloud Data Processing Addendum, with Google-approved Zero Data Retention. In plain terms, that is a special program ClassLens had to be approved by Google to use. Once approved, the AI model does not keep, store, train on, or log any of the student work that passes through it. Google does not keep it either. There is no prompt logging anywhere in the pipeline. The model is multimodal. That is why it can read a hand-drawn graph as readily as a paragraph. The reason it works for every subject, though, is the wrapper around the model. It is built to understand that a chemistry lab is not a five-paragraph essay and a CAD drawing is not a sentence with a verb in the wrong place.
You are still the teacher
ClassLens used to have an auto-return feature. The AI graded the work and sent the grade and comment straight to the student without another click from me. I built it because I was chasing efficiency.
I removed it in April 2026. The feature did exactly what I designed it to do, and I came to believe it was doing too much. Grades were reaching students without my eyes ever touching them, and that stopped sitting right with me. The longer story is for another day.
Now every grade lands in a Batch Review Dashboard first. You read what the AI suggested, change anything you want, and release the batch to your students when you are ready. You are the last set of eyes on every grade. Always.
That rule came out of my own classroom. It is non-negotiable in the product.
Try it
If you teach a subject where students turn in anything other than essays, you have probably already concluded that AI grading is not for you. The tools you tried were built for one subject. ClassLens is an AI-assistive grading tool built for every subject.
Sign in with Google. Pick an assignment. Set a rubric or paste one in. Let it draft the grades. Review the batch. Release when you are ready.
Try ClassLens at classlens.com
ClassLens is an AI-assistive grading and teaching tool built by Evolved Academics, LLC. Steven Swanson is a 22-year classroom teacher in California. He teaches engineering (design/drafting, mechatronics, and senior capstone) in the four-year engineering academy at Whittier High School, and AP Computer Science and AP Physics online. He built ClassLens after two days of chaperoning field trips produced 450 ungraded assignments and none of the tools he tried could grade them.