Founder @Babelize • Founding Engineer @ Tap In LLC
Full-Stack & AI/ML Engineer • Pune, India
I'm a Computer Engineering student in Pune who spends most of the week shipping production software.
I run Babelize, an AI localization platform for React, Next.js, Vue, Django and static sites. It scans a codebase for hardcoded strings, translates them with context, and opens a pull request with clean JSON locale files. Fifty plus languages, reachable over CLI, API, SDK or MCP. The whole thing started because I was tired of syncing translation keys by hand.
I'm also founding engineer at Tap In, a Brazilian Jiu-Jitsu app that pulls academies, open mats, class schedules and gym management into one place.
The rest of my time goes to AI/ML and IoT work: the kind that wires a model, an API and a physical device into one system that holds up.
name : Mohit Sapat
role : Founder @Babelize (AI localization platform)
Founding Engineer @ Tap In LLC (BJJ app for athletes & gyms)
education : B.Tech Computer Engineering, Vishwakarma University, Pune
location : Pune, Maharashtra, India
focus : Full-Stack Development, AI/ML, Computer Vision, IoT
learning : LLMs and RAG, Distributed Systems, Full-Stack Architecture
achievements: Published in Asia Inc. 500 magazine, May-June 2026
2x Hackathon Winner
Cricket Runners-up, HM. M.K. Karandak 2020
open_source : 10 PRs merged into billingsdk, DXLander and OSS Wiki
writing : medium.com/@sapatmohit18, mohitsapat.substack.com
orcid : 0009-0008-6057-6067
research : 5 papers, IEEE IGNITE 2026, IEEE ESCI 2025, ICSIAIML 2025, TANZ Vol 11
fun_fact : I build things that bridge the physical and digital world- Babelize: AI localization that scans your repo, translates in context and ships the change as a pull request
- Tap In: founding engineer on a BJJ app for athletes, academies and coaches
- Babelize Elements: open-source localization UI components for React and Tailwind
- Open source: 10 PRs merged into Dodo Payments' billingsdk, DXLander and OSS Wiki
Frontend
Backend & APIs
Databases
AI / ML & Data
Mobile, Hardware & IoT
Cloud, DevOps & Tools
|
Copy-paste language switchers, locale pickers and RTL layouts. Works with next-intl, i18next, react-intl, Lingui and FormatJS. MIT licensed.
|
AI localization platform. Scans your codebase for hardcoded strings, translates them in context and opens a pull request with clean JSON locale files. 50+ languages over CLI, API, SDK or MCP.
|
|
RAG advisory agent that turns live agronomic data into farming guidance a grower can act on.
|
Monitoring and control dashboard for industrial water pumps, wired to live hardware.
|
|
Production website for a non-profit, covering content, donations and outreach.
|
Hackathon builds, coursework that outgrew the assignment, and a few experiments that are still cooking. 
|
10 PRs merged across billing infrastructure and docs, plus a component library I maintain in the open.
|
Modern billing and monetization UI component library.
Contributor. Shipped the Billing Summary Card, Billing Settings, Upcoming Charges and Detailed Usage Table components. |
AI-powered zero-configuration deployment platform.
Contributor. Added Groq and OpenRouter provider support, per-config build deletion and custom encryption keys. |
|
Localization UI components for React. Language switchers, locale pickers and RTL layouts, MIT licensed.
Maintainer. I built the library, the docs site and the release flow, and I review and merge what comes in from contributors. |
Open-source wiki teaching people how to contribute to open source.
Contributor. Built the contributor tracker for documentation pages. |
I write about building products solo, burnout, and the parts of engineering nobody puts on a resume.
Published in Asia Inc. 500, May-June 2026. My essay From Building Things That Expire to Building Things That Last ran on pages 26 to 27 of the print issue, on how Babelize started and why I stopped building things with an expiry date.
- Developer Life: The Good, The Bad, and The Burnout · Jul 2026
- From Building Things That Expire to Building Things That Last · May 2026 · published in Asia Inc. 500
- The Pressure to Do Something Big · Apr 2026
- I'm Not Confident. I'm Just Continuing Anyway. · Apr 2026
- When Free Time Doesn't Feel Like a Break Anymore · Apr 2026
- The Week Babelize Almost Broke · Mar 2026
- The First Time My Code Actually Mattered · Mar 2026
- The Day I Stopped Building Things That Expire · Mar 2026
Five papers across applied machine learning and mathematics: computer vision, environmental modeling, medical imaging and Fourier analysis.
|
IEEE IGNITE 2026, Pune · pages 1-6 Aditya Bhope, Jameel Ansari, Mrunmai Ranade, Sandhya Tapadia, Mohit Sapat Works through Fourier series from first principles: coefficients, convergence and orthogonality, with full derivations. Then puts the theory to work in computational simulations, with a focus on signal processing and pattern generation.
|
IEEE ESCI 2025 · pages 1-7 Mohit Sapat, Sarthak Wadegaonkar, Sarthak Patankar, Pratibha Mahajan Benchmarked three deep learning models on the Indian Food Classification Dataset and Food-11 across accuracy, precision, recall and F1. InceptionV3 took the accuracy crown, while YOLOv8n held up best in real time, which makes it the one you actually ship to mobile or embedded hardware.
|
|
ICSIAIML 2025, Atlantis Press · pages 182-194 Maya Kurulekar, Mohit Sapat, Richa Panchgaur Compared Linear Regression, Random Forest and XGBoost for predicting the Water Quality Index of Indian rivers. XGBoost came out ahead at R2 0.89, RMSE 5.1, MAE 3.7, and SHAP analysis surfaced BOD, COD and DO as the parameters that move the prediction most. Built for evidence-based water governance under SDG 6.
|
Applying Theoretical Approach Towards Deep Learning Based Feature Fusion to Multimodal Medical Imaging Data for Improved Bone Cancer DetectionTANZ Research Journal, Vol 11 Issue 4, 2025 · pages 650-656 Purva Thakare, Mohit Sapat Fuses features across imaging modalities so a bone cancer detector reads more than one view of the same patient, rather than treating each scan type on its own.
|
Early Detection of Multiple Cancer Types Using Hybrid Machine Learning and Natural Language ProcessingTANZ Research Journal, Vol 11 Issue 4, 2025 · pages 313-320 Purva Thakare, Mohit Sapat Pairs classical machine learning with NLP over clinical text to flag several cancer types earlier than single-signal screening manages on its own.
|
Reviewer 2 · pages ∞ Mohit Sapat & a very patient inbox A few more papers are currently stuck in peer review, quietly aging like fine wine (or milk, still undecided). Status: "Under Review" since roughly the last ice age. Will update this card once Reviewer 2 wakes up.
|
__%2F%C2%AF-E0AF68?style=flat-square&labelColor=1A1B27){kind=icon}
"There's never a right time. Right now is the only right time."