Amit Yadav

What it is

A multi-agent AI pipeline that turns a single text prompt into a cohesive, 60-second cinematic video — orchestrating 10 specialized agents that handle scriptwriting, storyboarding, parallel rendering, and final stitching, with a built-in human-approval checkpoint before final generation.

Problem it solves

Advanced video generators like Omni Flash cap out at 10-second clips, blocking long-form storytelling. Manually chaining short clips together creates jarring visual seams — characters, lighting, and environments shift between shots — and creators waste hours bouncing between separate tools for writing, storyboarding, generating, and editing.

How it solves it

Dedicated agents lock in character design and cinematography rules before rendering, then 6 Omni Flash agents render six 10-second clips in parallel and stitch them into one continuous 60-second movie with identical style across scenes — compressing a multi-hour production pipeline into under 3 minutes.

What it is

A full-stack GenAI application that gives structured, honest, ATS-calibrated feedback on a resume for a target engineering role, with streaming output, strict input validation, anti-hallucination prompt design, and request-level observability.

Problem it solves

Generic resume feedback tools give vague, one-size-fits-all advice or require a human reviewer. Engineers switching roles or levels need fast, specific, honest feedback calibrated to their actual experience and target role, not a canned checklist.

How it solves it

A role- and experience-aware system prompt drives Gemini to stream a structured JSON review (summary, strengths, gaps, missing ATS keywords, suggestions, an honest 1-10 score) over Server-Sent Events, with anti-hallucination rules, full input validation before any model call, and every request logged with token usage, latency, and outcome.

What it is

A multi-city gym-booking web app that lets users schedule workouts across partner gyms based on availability and location.

Problem it solves

Fitness enthusiasts who travel or split time across cities have no single place to discover partner gyms and book a session at a specific time slot, backed by real payments and identity verification.

How it solves it

Integrated Google OAuth 2.0, phone OTP verification, Razorpay payments, Google Maps API, and Cloudinary media uploads, all backed by a normalized MongoDB schema and Node.js/Express APIs.

What it is

A responsive shelter-home platform supporting pet adoption, volunteering, emergency assistance, and shelter visits, with an accessibility-first UI.

Problem it solves

Animal shelters need an accessible, single platform where adopters, volunteers, and donors can discover pets, sign up to help, and contribute — instead of scattered spreadsheets, phone calls, and social posts.

How it solves it

A full-stack React, Node.js, Express, and MongoDB architecture with PayPal integration for secure donations, deployed to Vercel with environment-scoped CI to keep staging and production isolated.