Full-stack engineer based in Surat, India — AI-native by default. I architect distributed systems, migrate terabytes of data with zero downtime, and ship production-ready platforms — from Next.js monorepos powering 12+ brands to ML models running at 92% accuracy. Available for freelance and full-time roles.
For 2+ years I've been building full-stack systems that run at scale — from a 35TB cloud migration pipeline with zero downtime to a Next.js monorepo powering 12+ brand websites, to a diamond e-commerce backend processing real-time ERP data. Based in Surat, India, I obsess over the seam between a system's architecture and the product it powers.
I work in TypeScript, Next.js, Node.js, and Python. I design distributed data pipelines and manage cloud infrastructure on GCP and AWS. I've shipped OpenSearch integrations, Pimcore DAM/PIM/MDM systems, and Medusa.js backends end-to-end. I've also built ML models achieving 85–92% accuracy and integrated LLM APIs into production SaaS products. I'm currently open to freelance projects and full-time opportunities — remote or based in India.
Each role earned the next. The camera holds on each chapter for a beat — scroll to advance.
Built a full-stack B2B e-commerce platform for the diamond industry. Improved performance by 40% through lazy loading, browser caching, image compression, and efficient MongoDB queries.
Researched AI/ML-driven smart building systems. Built predictive models in Python and TensorFlow for IoT-enabled automation, achieving 85% prediction accuracy.
Developed a multi-tenant School ERP system with Next.js, TypeScript, and PostgreSQL. Implemented role-based auth, custom domain/subdomain routing, and modular workflows.
Engineered a Next.js monorepo for 12+ brand websites and architected a 35TB cloud data migration with zero downtime. Also built a scalable diamond e-commerce backend with Medusa.js.
Tech stacks tell you what someone has touched. The bars below tell you what they can actually do.
Vertical scroll moves the panel horizontally — the page is the camera, the work is the scene.
Unified Next.js monorepo for 12+ brands. 35TB cloud migration, Pimcore DAM/PIM/MDM, GCP infrastructure — zero downtime.
Production Medusa.js + PostgreSQL backend with queue-driven ERP ingestion, OpenSearch, and Strapi CMS for zero data lag.
Full-stack real estate SaaS. 92% ML valuation accuracy, LLM APIs, Google Maps, Stripe subscriptions, 500+ listings.
Django REST + NLP + LLMs + TF-IDF. Parses PDF/Word up to 50MB, Google Search API web matching, 90% accuracy.
Multi-tenant Next.js SaaS with RBAC, custom subdomain routing, and 4 modular workflows built at AppArrow Infotech.
Drag the handle. The architecture diagram on the left, the shipped product on the right. Same vision, same data models, same eyes.
Real estate platforms lacked intelligent valuation — agents priced manually, buyers had no transparency, and financial tools were disconnected from actual market data. Existing SaaS solutions were monolithic, hard to scale, and lacked AI-driven insights.
Built a microservices SaaS with TensorFlow ML models trained on real property data achieving 92% forecasting accuracy. Integrated LLM APIs for natural language loan estimation and automated content generation. Added Google Maps for geospatial property insights, Stripe for subscription billing, and a multi-tenant PostgreSQL architecture for secure data isolation across 500+ listings.
92% ML forecasting accuracy on property valuations. 500+ listings managed with full multi-tenant database isolation. LLM-powered loan estimator and NLP content generation live. Stripe subscription billing automated end-to-end. Full-stack from model training to deployed production UI on GitHub.
Four steps. Same shape every time, different details every project.
I read requirements like a spec — every data relationship, every scale constraint. I write down the hard questions before I write code.
Service boundaries, data models, infrastructure plan. I'd rather spend a day here than three days untangling a distributed mess later.
Idempotency, checkpointing, graceful degradation. Every pipeline has a recovery path — production systems don't get second chances.
Zero downtime deployments. Performance benchmarks enforced. Cloud costs monitored. The reliable version is the only version.
Move your cursor. Eighteen-by-ten lines, each pointing toward you, weighted by distance. Sixty lines of canvas, ~120 lines of math, zero dependencies.
Every section on this page has at least one of these — a small interaction that earns its place.
Sourced from teammates across four companies — names redacted as placeholders until I have written permission to publish.
Meet has the instincts of an architect and the discipline of a senior engineer. The 35TB migration he led hit every milestone with zero downtime.
I gave Meet requirements and he gave back a running system. The School ERP he built handles multi-tenancy, RBAC, and custom domains — all in one clean codebase.
Ships fast, ships clean, and the engineers we have in-house come away from every collab having learned something.
Selective collaborations. Two-week minimum, three-month sweet spot. I reply within a working day.
meetvaghani1239@gmail.com[ click to copy ]