Skip to content

Production work ยท Case studies

Selected Work

Multi-LLM systems, RAG pipelines, document automation, and full-stack products I have helped design and ship.

Context and confidentiality

Much of my production work belongs to employers or clients, so I do not publish private repositories, credentials, customer data, or internal screenshots. The summaries below focus on the problem, my engineering responsibility, the stack, and measurable operating context that can be shared publicly.

Professional experience

Software Engineer

Varahe Analytics Pvt. Ltd. ยท Noida, India

  • Architected and built full-stack AI platform from the ground up โ€” owning frontend, backend, and cloud infrastructure end-to-end.
  • Engineered a unified multi-provider AI abstraction layer routing text and image generation across multiple LLM providers with runtime model switching and dual-key failover.
  • Built RAG pipelines using Pinecone vector database for semantic style retrieval, script similarity search, and artist-style matching across namespace-isolated indexes.
  • Developed an AI-powered entity recognition pipeline using Gemini โ€” auto-detecting named subjects from user prompts and injecting reference images into AI generation workflows.
  • Integrated ElevenLabs TTS and Gemini multimodal translation for content generation across 9 Indian languages; implemented i18next for frontend multi-language switching.
  • Built admin analytics dashboard using BigQuery and Recharts for usage tracking, AI spend monitoring, and automated PDF report generation.
  • Implemented secure authentication with Google OAuth, JWT in httpOnly cookies, Helmet headers, CSRF protection, and rate limiting.
  • Temporarily led a cross-functional engineering team โ€” conducting code reviews, architecture planning, and sprint management across multiple AI feature deliveries.
  • Designed and shipped AI image-tagging pipelines, content-scraping automation, cartoon generator, and infographic generation engine.

Junior Developer

Yatri Restro ยท Patna, India

  • Built the company's flagship web platform using Next.js and React Query with an integrated online payment gateway.
  • Developed a custom admin dashboard in React, Node.js, and MongoDB for order management and business analytics.

Production case studies

Production system

AI Content & Automation Platform

Full-stack AI content generation platform producing scripts, posters, infographics, thumbnails, cartoons, and voiceovers across 9 languages from a single interface. Implemented a multi-provider AI abstraction layer with runtime model switching, alongside a RAG pipeline using Pinecone for style retrieval. Features Algolia-powered searchable media galleries (140K+ assets), BigQuery analytics dashboard, and animated template export with HTML/GIF output. Deployed on GCP Cloud Run via Docker with Cloud Build CI/CD to support high-throughput production workloads.

Stack: React.js, FastAPI, Express, MongoDB, Gemini, OpenAI, Pinecone, Algolia, GCP, Docker

Production system

Bulk Document Generation Suite

Bulk PDF generator producing 500+ personalized documents (ID cards, certificates) in under 1 minute โ€” replaced a full day of manual design work. 4 secure internal dashboards with RBAC โ€” Budget, Vendor Management, Internal Tracking, and real-time analytics.

Stack: React, Node.js, MongoDB, PDF Rendering, GCP

Production system

Bulk ID Card Generator

Generates 100โ€“500+ ID cards instantly from dynamic Google Sheets data โ€” reduced a full day of manual design to under 1 minute.

Stack: React, Node.js, MongoDB, PDF Rendering

Creator workflow: 9 Face Toon

Outside software work, I run a Hindi animation channel with an audience of more than 40K on both Instagram and YouTube. It is where I test short-form production workflows, animation tools, captions, voice, thumbnails, and the practical limits of AI-assisted creative work.

View the public work on Instagram and YouTube, or read the channel growth case study.

Earlier learning project

Team learning project

ShopDressUp

Led 3-member team to ship a full-stack fashion marketplace in 5 days โ€” product filtering, cart, checkout, and fully responsive UI.

Stack: React, Node.js, Express.js, MongoDB, Chakra UI

Discuss a project

If you need a full-stack product, AI integration, RAG workflow, or automation system, send the problem, expected users, and current stack.