Engineer.Builder.Storyteller.
Building products where software, AI and storytelling meet.
India · Working with teams worldwide
About
Building software people
actually enjoy using.
I build products across AI, software and creative technology.
From full-stack applications and computer vision systems to cinematic AI workflows, I enjoy turning ambitious ideas into things people can actually use.
Shipped Projects
Luxury Commerce • Fitness AI • Film
Industries
E-Commerce • Health-Tech • Entertainment
Core Technologies
AI • Full Stack • Computer Vision
Selected work
Bespoke digital products.
Engineering and art direction combined — full-stack commerce platforms, AI-powered generation engines, computer vision systems, and cinematic production workflows.
Daddy Prince
A luxury home décor and heritage commerce platform engineered for high performance, secure payments and a premium shopping experience.
Daddy Prince curates premium paintings, heritage art and home décor. I designed and engineered the storefront from scratch, focusing on performance, secure checkout and an immersive luxury shopping experience.
What I Built
Engineered a custom Next.js storefront featuring an editorial luxury UI, secure user authentication, and full Razorpay integration. Focused heavily on performance optimization and dynamic commerce interactions to deliver a high-end client experience.
Technical Details
Leveraged Next.js App Router and Supabase for a scalable server-side architecture. Implemented Row-Level Security (RLS) policies to protect customer transactions, built server-side webhook signature checks for secure payment validation, and configured optimized image pipelines.
Results
- Funnel checkout flow reduced to three steps.
- Optimized responsive image pipeline loads in under 1.2s.
- Secured customer accounts using Supabase row-level policies.
TakeawayPayments teach you defensive programming. Verifying signatures and handling edge-case checkout drops took more work than the happy path, showing why failure states come first.
Forge
Describe a website layout in plain English. Forge generates it as editable React component trees.
Forge is an AI generation engine that compiles natural language prompts into React components. Unlike static code screenshot tools, the output remains fully editable.
What I Built
Structured the model outputs as components rather than raw HTML blocks. Created a sandbox preview iframe that syncs dynamically as the node tree is adjusted.
Technical Details
Implemented LLM structured JSON schemas to plan layout hierarchy before code writing. Built prompt validation rules matching a strict component catalog.
Results
- Stable React component compiler assembly completed.
- Live sandbox preview rendering synchronizes in real time.
- Zero-dependency component outputs fit standard Tailwind setups.
What I Built
Integrated MediaPipe to track 33 key body landmarks. Wrote a joint-angle state machine that tracks movement phases and ignores incomplete repetitions.
Technical Details
Calculated exercise coordinates on camera feeds using OpenCV and Python. Connected an LLM assistant to generate custom nutrition plans from tracking history.
Results
- Maintains repetition accuracy across tempos and incomplete reps.
- Runs completely locally in-browser with zero server inference cost.
- Designed layout to support HTML5 video upload file processing.
FitHub AI
Webcam pose estimation that tracks movement, counts repetitions, and corrects exercise form locally.
FitHub AI reads standard webcam feeds to count repetitions and check joint postures. The application runs locally in the browser with no external hardware required.
GDN Title Reveal
Designed for a feature film.
Currently under release embargo.
Full breakdown available after the official release.
AI Web Scraping & News Discovery Agent
Custom agentic system that scrapes web targets, sanitizes HTML data, and extracts trending technical updates.
Wrote an autonomous agent to scrape target documentation sites, clean HTML blocks, and expose active tech newsletter newsletters.
What I Built
Orchestrated cron pipelines connecting internal target scrapers with visual filters.
Technical Details
Deployed serverless event handlers with row-level database sync and Slack logging loops.
Results
- Automated intake processes from custom sources.
- Fails loudly on scraper target blockages.
TakeawayScraping pipelines require designing for failure first. Handling target IP blockages, parsing dynamically-rendered HTML, and sanitizing LLM payload size limits taught me to program defensively.
Capabilities
Things I enjoy building.
AI that behaves predictably in production.
I design structured model outputs, prompt architectures, and local agents with robust recovery handlers.
Process
How the work actually happens.
Think
Map parameters, identify constraints, and scope project outcomes.
Prototype
Build rapid clickable flows early to test mechanics and layouts.
Build
Develop database schemas, secure routes, and clean frontend components.
Polish
Refinement phase for typography sizes, loading states, and cursor interactions.
Ship
Deploy on cloud hosts with performance tracking and error logging.
Playground
Experiments and learning.
Colour · Look Development
Grade studies
Establishing camera lens styles, lighting angles, and colour grading before starting renders.
Sequencing · Pre-Viz
Storyboard passes
Sequencing beats first so visual generation matches the edit timing accurately.
— anamorphic 40mm, shallow depth of field
— single hard key from frame left, deep falloff
— muted teal shadows, warm sodium highlights
— fine 35mm grain, no digital sharpening
— subject: {SHOT_SUBJECT}
Prompt Engineering
Prompt schemas
Fixed base rules and variable slots to build predictable generative pipelines.
Type · Motion
Kinetic typography
Typography animation reveals, mask offsets, and reading-rhythm easing curves.
Interaction · Motion Design
Motion grammar
Unified UI movement guidelines to maintain clean navigation transitions.
Branding
Identity explorations
Brand identity grids, wordmark sizing, and logo baseline geometry studies.
Tech Stack
The technical
spec sheet.
A toolbox shaped by real projects. No speculative tools, no checklist extensions — only what is actively deployed, tested, and maintained.
- Next.js
- React
- TypeScript
- Tailwind CSS
- HTML
- CSS
- JavaScript
- Node.js
- Python
- FastAPI
- REST APIs
- Supabase
- Authentication
- PostgreSQL
- Supabase
- SQL
- OpenAI
- Gemini
- Claude
- MediaPipe
- OpenCV
- Prompt Engineering
- LLM APIs
- AI Automation
- AI Agents
- ComfyUI
- Image Generation
- Workflow Automation
- ComfyUI
- Midjourney
- Runway
- Seedance
- Kling
- ChatGPT
- Claude
- Gemini
- Cursor
- Antigravity IDE
- Vercel
- Render
- GitHub
- Supabase
Experience
Not a résumé. A trajectory.
No two years here look the same, on purpose — exploration first, then increasing weight with every project since.
The starting line
B.Tech, Computer Science — AI specialisation
Began my degree and spent the year genuinely exploring: programming fundamentals, a scattering of technologies, no fixed direction. Looking back, the useful thing about this period is that I wasn't pretending to know yet.