You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$250,000 left to reach the funding goal.
I am a mathematics researcher working in finite group theory. Like other mathematicians, I write my research first on my notebook and than type it in overleaf.
Turning these handwritten notes into a verified research paper is time consuming. We have to structure the proofs, check for grammatical mistakes, shorten the proofs and write references with a nice introduction and title. All of these produces no new work and we cant even use ai directly, because unconstrained prompt will lead to hallucinations and wrong proofs and wrong structure of the whole paper. This kills a lot of time on daily basis, the time which can be used for other creative work.
I built Scriptavera exactly to fix this. It is an open source 20 step pipeline under AGPL 3.0 license with over 15,000 lines of functional Python and JavaScript already public on GitHub. Instead of letting an AI guess formulas, a mechanical formula auditor (audit.py) converts both the notes and the LaTeX into abstract syntax trees to guarantee that every equation matches the handwritten notes symbol by symbol. It verifies citations across five academic databases and pauses at six human checkpoints so that the mathematician remains in control during the whole process.
It also integrates RFC 3161 cryptographic timestamps. In pure mathematics, priority is everything. Mathematicians often worry that there ideas can be stolen if they will upload there notes ont the internet or could have disputes over who proved a lemma first. To solve this exact problem we use cryptographic signature. When you run your notes through our software Scriptavera, it seals the cryptographic hash of your original handwritten scans alongside the final draft with a verified digital timestamp authority. This gives you legally recognization that you wrote this proof first.
I am raising 250,000 USD to secure a two year runway for our core engineering team. Raising two years of runway solves the biggest problem with grant funding which is job security for hired engineers. It lets me hire two senior engineers on multi year contracts without yearly renewal cliffs, giving us the stability to harden the audit engine, add formal proof assistant integration, and deploy the system across the mathematical departments of various institutes.
Our primary goal is to eliminate the massive friction researchers face when turning handwritten notes into publication quality research papers.
Mathematicians and scientists do their original thinking on paper, but manually typing handwritten notes, proofs, and formulas into Overleaf takes weeks of tedious work that pulls researchers away from creative discovery. Scriptavera automates this entire path so authors can go directly from physical notes to finished research papers within minutes instead of days or weeks.
Specific Project Goals:
1. Friction Free Paper Drafting: Transform handwritten mathematical notes directly into complete, publication ready research papers with verified bibliographies and clean Overleaf export.
2. Symbol Level Faithfulness: Guarantee that every formula in the generated paper matches the handwritten notes symbol by symbol using deterministic abstract syntax tree auditing in audit.py.
3. Priority and Authorship Protection: Use RFC 3161 cryptographic timestamps to seal handwritten source notes to the final draft, giving researchers a proof of discovery date before public distribution.
4. Open Scientific Infrastructure: Release all code freely under AGPL 3.0 so students and researchers globally have free access to automated paper writing tools.
How We Will Achieve Them Across Our Two Year Plan (250,000 USD):
Year 1 (100,000 USD): Mathematics First, Backend Hardening, and Pilot Testing
We tackle pure mathematics first because it has the strictest notation. I will hire a dedicated senior backend engineer and an AI engineer to turn our 15,000 line functional prototype into robust production software. We will further improve the 20 step pipeline using Docker, implement Celery and Redis worker queues for asynchronous document handling, and harden the mechanical formula auditor. We will integrate RFC 3161 digital timestamping and run pilot testing with university mathematics faculty to refine the user experience on real research drafts.
Year 2 (150,000 USD): Full University Deployment and Expansion to Other Scientific Disciplines
Our main technical milestone in Year 2 is expanding Scriptavera beyond mathematics to other formula heavy disciplines like theoretical physics, computer science, and engineering. We will deploy high throughput GPU clusters to serve multiple university departments concurrently. Finally, we will launch hosted institutional tiers for university departments, creating recurring subscription revenue to make the project permanently self sustaining.
We are raising 250,000 USD to fund our two year plan. Year 1 matches our master budget in funding.json (100,000 USD), while Year 2 (150,000 USD) funds multi university deployment and expansion into physics and computer science.
Year 1 Budget Breakdown (100,000 USD):
1. Personnel (75,000 USD):
Vaibhav Chhajer (Principal Investigator and Mathematical Lead): 25,000 USD for twelve months full time mathematical direction and pipeline architecture.
Senior Full Stack Backend Engineer: 25,000 USD for twelve months full time systems engineering, Docker containerization, and Celery worker queues.
Applied AI Engineer: 25,000 USD for twelve months full time OCR extraction, abstract syntax tree formula verification in audit.py, and model routing.
2. Materials and Testing Hardware (5,500 USD):
Three developer workstations, high resolution external monitors, and an iPad with stylus for digitizing and testing real handwritten mathematical notes.
3. Cloud GPUs and Frontier AI APIs (10,000 USD):
Dedicated GPU server rental on RunPod or Lambda and API compute credits to run the 20 step pipeline across thousands of test pages.
4. Operations and University Testing (9,500 USD):
Dedicated three seat coworking cabin with high speed internet: 5,500 USD.
Domestic travel across university mathematics departments for live faculty pilot testing: 2,000 USD.
Legal compliance, accounting, and RFC 3161 cryptographic priority timestamping tokens: 2,000 USD.
Year 2 Budget Breakdown (150,000 USD, Deployment and Scientific Expansion):
1. Expanded Engineering Team (104,000 USD):
Senior Full Stack Backend Engineer: 32,000 USD.
Applied AI Engineer: 32,000 USD.
Vaibhav Chhajer (PI and Mathematical Lead): 28,000 USD.
Junior QA and Benchmark Engineer: 12,000 USD to run continuous test suites as we expand the pipeline to physics and computer science.
2. Scaled Cloud GPUs and Model APIs (22,000 USD):
Multi GPU clusters and expanded API capacity to serve concurrent document processing across dozens of university departments without delays.
3. Testing Hardware and Lab Equipment (6,000 USD):
Upgraded developer hardware and multi device test bench for multi page document processing.
4. Field Expansion, Workshops, and Operations (11,000 USD):
Travel for expanded nationwide university cohorts and faculty onboarding workshops: 4,000 USD.
Managed four seat office cabin and enterprise internet: 5,500 USD.
Legal compliance, reporting, and RFC 3161 priority tokens: 1,500 USD.
5. Contingency Buffer (7,000 USD):
Standard five percent reserve to protect against cloud compute cost spikes and currency fluctuations.
What We Accomplish with Minimum Funding (25,000 USD):
The minimum funding of 25,000 USD secures our Quarter 1 milestone. It allows me to immediately hire our senior backend engineer on a three month contract to containerize the 20 step pipeline using Docker, implement Celery and Redis worker queues for heavy document processing, harden the AST formula auditor in audit.py, and integrate RFC 3161 cryptographic priority timestamps. This delivers a production grade engine ready for initial faculty testing.
Who is working on this:
Vaibhav Chhajer (Founder and Mathematical Lead):
I am an independent mathematics researcher and the creator of Scriptavera. I finished my Master of Science in Mathematics at the Central University of Karnataka, scored All India Rank 299 in the GATE 2024 mathematics exam, and spent a year doing pure math research at NISER Bhubaneswar. My work is in finite group theory.
Here are my research papers:
1. Published in the Indian Journal of Pure and Applied Mathematics (Springer):
Finite groups with mostly involuted cyclic subgroups (with Palash Sharma)
DOI: https://doi.org/10.1007/s13226-026-01029-y
arXiv: https://arxiv.org/abs/2508.00681
2. Accepted in Proceedings Mathematical Sciences (Springer and Indian Academy of Sciences):
Finite groups with nearly half as many cyclic subgroups as elements (with Sumana Hatui and Palash Sharma)
arXiv: https://arxiv.org/abs/2506.21163
I wrote the current codebase myself from scratch. It is over 15,000 lines of working Python and JavaScript on my GitHub:
https://github.com/Chhajer2000/scriptavera
The people I will hire with this grant:
1. Senior Full Stack Backend Engineer (Year 1 and Year 2):
This engineer will help me improve the 20 step pipeline with Docker, manage Celery and Redis worker queues for heavy PDF processing, and set up RFC 3161 cryptographic priority timestamps.
2. Applied AI Engineer (Year 1 and Year 2):
This engineer will focus on tuning local vision models, refining equation parsing in audit.py, and helping expand the tool to physics and computer science notes.
3. Junior QA and Benchmark Engineer (Year 2):
This person will run test suites and check that real manuscripts from partner universities compile cleanly without formula errors.
My track record:
1. Building Scriptavera from scratch:
I built this entire project alone so far. It is a working tool with thousands of lines of open source code under the AGPL 3.0 license. I built the full 20 step pipeline, the AST formula checker (audit.py), the citation lookups across five databases, and the six author verification checkpoints. You can see me running the software on my own machine in this two minute video:
2. Publishing real mathematics:
I am not building this from the outside as a generic software developer. I am an active pure mathematician who deals with this problem every day. I know how mathematicians work, how proofs are written and how much time consuming it is to write a research paper. Having experienced the problem myself i decided to build this software.
Here is an honest assessment of how this project could run into trouble and what happens in each case.
Most Likely Causes of Failure:
1. Messy Handwriting and Extreme Notation Edge Cases (Technical Risk):
Some mathematicians write with terrible handwriting or personalized shorthand symbols. If notes are too degraded, OCR extraction can produce low confidence segments.
How we handle this:
Scriptavera does not attempt black box guesses. It stops at six author checkpoints. When a symbol or figure is unclear, it flags the image segment and asks the author directly instead of guessing. Even if full automation fails on very messy pages, the human in the loop checkpoints keep the output mathematically correct.
2. Academic Habit and Faculty Inertia (Adoption Risk):
Senior professors can be slow to change their habits. Many senior faculty are used to typing directly into Overleaf or suggest manual typing to their PhD students.
How we handle this:
We are focusing our pilot testing on PhD students, postdocs, and junior researchers. These are the people who actually do the heavy scratch derivations on paper and spend nights typing formulas into LaTeX. They feel the pain of lost research time most acutely.
3. Engineering Hiring Delays (Execution Risk):
Finding sharp backend engineers who understand both distributed systems and mathematical abstract syntax trees can take time.
How we handle this:
We offer competitive market compensation and a clean, functioning codebase on GitHub. The fact that the code is open source and solving a real scientific problem makes it much easier to attract engineers who care about open science.
What Happens if the Project Fails (The Outcome):
Even in the worst case scenario where we fail to achieve widespread multi university adoption, this project is not a total loss.
Because Scriptavera is released openly under AGPL 3.0, every line of code we write belongs to the scientific public. The mechanical formula auditor (audit.py), the 20 step conversion architecture, the citation validation tools, and our handwritten test benchmarks will remain permanently free and open source on GitHub. Other researchers and open source developers will be able to take our audit engine, fork it, and use it to keep AI transcription honest.
Zero outside funding. I have raised 0 USD in the last twelve months.
Scriptavera is 100 percent self funded and bootstrapped to date. I built the working pipeline, the formula auditor, and the web interface using my own personal savings, hardware, and time.
To fund our upcoming two year engineering roadmap, we have submitted applications to four non dilutive grant programs over the past week, which are currently under active review:
1. AI for Math Fund Seed Grant: 99,000 USD (Pending review)
2. Foresight Institute AI for Science Nodes: 99,000 USD (Pending review)
3. Effective Altruism Funds Transformative AI Fund: 150,000 USD (Pending review)
4. The RAAIS Foundation Open Source AI Grant: 25,000 USD (Pending review)
We have received no funds from these or any other external sources yet.