Divyansh Lalwani
Divyansh Lalwani (known professionally as Dev Lalwani) is an Indian-American software entrepreneur and biomedical engineer who founded LayerNorm in February 2026.11 He studied Biomedical Engineering and Applied Mathematics and Statistics at The Johns Hopkins University, graduating in December 2025.1 LayerNorm's flagship product, Overlay, is an open-source AI workspace that Lalwani positions as the open interface layer for AI — the role the web browser played for the internet.7
Early life and education
Lalwani grew up in Jaipur, Rajasthan, India, where he attended Jayshree Periwal International School. He scored a perfect 36 on the ACT, graduated as valedictorian, and served as student body president of a 2,100-member student body.1 During high school he was also a founding member or leader of more than ten student organizations, including The Artistic Syndicate, IMPACT (head of technology), and JPIS Council.1
In May 2021, during India's second COVID-19 wave, Lalwani co-founded Students for Unified Relief, an initiative that raised approximately $53,000 to supply oxygen concentrators and compressors to hospitals and health centers in both urban and rural Jaipur.1 The effort was recognized by the Chief Medical and Health Officer of Jaipur — the highest medical authority in the city — and covered by News18 Rajasthan and NewsNation.1
From September 2021 to May 2022, Lalwani participated in The Knowledge Society (TKS), a competitive innovator program for STEM-oriented students that covered artificial intelligence, quantum computing, brain-computer interfaces, and aging research. His cohort also consulted for IKEA on sustainability strategies.1 Concurrently, from November 2021 to January 2022, he worked as a machine learning intern at NeuroEquilibrium Diagnostic Systems in Jaipur, training deep convolutional neural networks on retinal OCT scans to classify diabetic retinopathy by severity. Using ResNet50 as the model architecture, he achieved 95.2% accuracy.21
Lalwani enrolled at The Johns Hopkins University in May 2022, studying Biomedical Engineering and Applied Mathematics and Statistics. JHU's BME program is ranked first nationally.1 He was active in the Biomedical Engineering Society, the Brain Computer Interfaces Society, and PharmHop, and served as a resident advisor beginning in 2023.1
Neuroengineering and biomedical research
Between February and July 2023, Lalwani worked as an undergraduate researcher in the Neuroengineering and Biomedical Instrumentation Lab (Thakor Lab) at the JHU School of Medicine. The project developed in-ear electroencephalogram (EEG) systems for use in animal models, wearable devices, and brain-computer interface applications. He designed electrodes from scratch, testing biomaterials to optimize for electrical conductance while minimizing impedance, and assisted in rat neurosurgeries.1
From January to August 2023, Lalwani also participated in the JHU Biomedical Engineering Design Team, developing AptaTech — an aptamer-based electrochemical assay for portable, point-of-care diagnosis of ischemic stroke and quantification of stroke severity. The project was advised by Dr. Robert Stevens, with the goal of integrating AptaTech into ICU workflows at The Johns Hopkins Hospital.1
Also in 2023, he built a computer vision system for motor rehabilitation at JHU Medicine that tracked joint motion in post-surgery patients, achieving 96.7% accuracy in rehabilitation analysis.2 In the summer of 2024, Lalwani interned as a statistical programmer at Bristol Myers Squibb. He built Python pipelines combining programmatic and optical character recognition to validate data in clinical trial report documents for FDA drug applications, achieving 100% accuracy against ground-truth data. The system saved an average of 20 hours per week for each of 18 human data checkers, eliminating approximately 360 person-hours per week from the validation process.1
AutoQuill
Lalwani began building productivity software in his junior year at JHU, connecting his interest in brain-computer interfaces with the goal of increasing the "bandwidth" of human interaction with computers.3 His first product, AutoQuill, is a macOS application that allows users to press a global hotkey from any application and dictate text that is transcribed and pasted into the active window.5
AutoQuill was built as a free, open-source alternative to WisprFlow, a paid voice dictation application.3 Lalwani published the app under the MIT license using the Flutter framework, targeting macOS 10.14 (Mojave) and later.6 The application calls Groq's API to run the Whisper speech recognition model, returning transcriptions in under 500 milliseconds.5 It supports three input modes: standard transcription via a single hotkey, push-to-talk by holding a key down, and an AI assistant mode that accepts a screenshot of selected text and voice instructions to edit or generate content.6
Lalwani announced AutoQuill to the Groq developer community on June 5, 2025.4 Unlike WisprFlow, which limits free users to 2,000 words per week, AutoQuill offers unlimited transcription with no account required; users supply their own Groq API keys, and all transcription data stays on the device.5 The GitHub repository accumulated 35 stars and 4 forks.6
Lalwani spent the summer of 2025 in San Francisco, where he described meeting people in the software startup ecosystem as decisive for his path as a founder.3 He was also named a finalist for Neo, a competitive fellowship for entrepreneurial undergraduates. Neo's alumni include founders of Cursor, Cognition, Pika, and Chai.101
LayerNorm and Overlay
Lalwani graduated from Johns Hopkins in December 2025 and founded LayerNorm in February 2026.111 The company's stated mission is "building the future of personal computer interfaces."11 The name LayerNorm references layer normalization, a mathematical operation used in transformer neural networks to stabilize training.
The company's main product, Overlay (distributed at getoverlay.io), is an open-source, web-based AI workspace that aggregates multiple large language models — including GPT, Claude, Gemini, Grok, and DeepSeek — into a single interface.8 Lalwani's published manifesto argues that AI models are becoming commodities while the interface layer is the actual locus of lock-in, and that an open aggregator can succeed by analogy to how Chrome gained market share from Internet Explorer and VS Code became more widely adopted than JetBrains-style IDEs.7
Overlay's free tier provides unlimited access to an auto-selected model and core workspace features. Paid tiers start at $8 per month (Starter), rising to $24 (Pro) and $94 (Max), with custom budgets available up to $200 per month; each paid tier unlocks premium models, browser automation tasks, image and video generation, and code execution via Daytona sandboxes.9 The web client's source code is published on GitHub under the repository DevelopedByDev/overlay-web.8
Lalwani described his progression from AutoQuill to Overlay in a March 2026 interview with the Johns Hopkins News-Letter: "My thesis then became, 'Okay, I've already made the way you communicate with the computer [to have] higher bandwidth. How do I make it so that when you use your computer, you do more useful work in less amount of time?'"3 Overlay initially launched as a macOS desktop application with four overlay panels — speech transcription, notes, AI chat, and a browser — that appear on top of the user's current screen and can be dismissed without fully switching context. The product later became available on the web at getoverlay.io, requiring no installation.8
References
- Divyansh Lalwani — LinkedIn(accessed Apr 24, 2026)
- Divyansh Lalwani — Personal Website(accessed Apr 24, 2026)
- AutoQuill: best (free forever and open source) voice-to-text app made with Groq — Groq Community(accessed Apr 24, 2026)
- AutoQuill — Transform Your Voice Into Perfect Text(accessed Apr 24, 2026)
- DevelopedByDev/autoquill_ai — GitHub(accessed Apr 24, 2026)
- The overlay manifesto — getoverlay.io(accessed Apr 24, 2026)
- Overlay — Open-source AI workspace(accessed Apr 24, 2026)
- Overlay pricing — getoverlay.io(accessed Apr 24, 2026)
- Neo Scholars — neo.com(accessed Apr 24, 2026)
- LayerNorm — layernorm.co(accessed Apr 24, 2026)