Abhishek Navgan
Pune, Maharashtra, India

Hi, I'm Abhishek Navgan

I build production-ready backend systems, automate cloud deployments,
and enjoy making infrastructure less painful.

Over the past few years I've worked across frontend, backend, AI, and DevOps, building products from idea to deployment. These days I'm focused on backend engineering and DevOps, while continuing to learn more about cloud infrastructure and understanding how the software reaches the end user.

Experience

Software Engineer Current

Nova Techset Ltd. · Pune

Full-time · Backend & DevOps

Freelance Full Stack & DevOps Engineer

Self-employed

SaaS apps · Cloud & deployments

Full Stack Developer / Team Lead

Somvanshi Technologies · Pune

Full-time · on-site

Full Stack Developer · Internship

XYZ Company · Pune

Full stack · React.js · Node.js

Technical Toolkit

03

AI & Integrations

Building AI applications using RAG, multimodal generation, and agentic workflows. I've worked with LLMs for document understanding, image generation, prompt optimization, and streamed AI interactions.

LLMs RAG Agents Multimodal Pinecone LangChain
04

Data Layer

Relational PostgreSQL · MySQL
Documents MongoDB
Cache Redis
Vector Search Pinecone
Tooling Prisma · SQLAlchemy · Alembic · Mongoose
05

Cloud Technologies

IAM Identity, access policies & role management
Compute VMs, containers & serverless workloads
Networking VPCs, DNS, routing & load balancing
Storage Object, block & file storage services
06

Frontend

Language
TypeScript
Frameworks
React · Next.js
Styling & UI
Tailwind CSS · HTML & CSS

Infrastructure Journey

Already Comfortable
  • Containerization
  • CI/CD Pipelines
  • Server Administration
  • Version Control
Currently Learning
  • Infrastructure as Code
  • Container orchestration
  • Observability
Next Up
  • Platform Engineering
  • Cloud Infrastructure
  • Monitoring & Reliability

Hobby Projects

Personal experiments which helped me learn and grow.

ThumbCraft

Generative AI thumbnail studio for YouTube-ready visuals. multimodal text-to-image and image-to-image generation with a multistep, AI-enhanced preferences flow. OpenAI prompt enhancement, 8 high-CTR SVG templates, direct Cloudinary CDN delivery, and searchable generation history with bulk ZIP export. Featured as a Top 3 project on ChaiCode's YouTube channel.

Generative AI Multimodal Multistep AI Cloudinary CDN Gemini OpenAI React Node.js
ECHO

Full-stack RAG document analysis platform for intelligent, context-aware Q&A over your own files. Upload DOCX, PDF, and more — documents are chunked, embedded, and indexed in Pinecone for fast retrieval. LangChain + OpenAI generate cited, source-backed answers from retrieved context. React + TailwindCSS frontend, Node.js + Express backend with rate limiting, CORS, and Google OAuth auth.

RAG LangChain Pinecone Vector DB OpenAI React Node.js MongoDB
Devnest

AI-powered full-stack learning platform for developers. RBAC course management with sections, video lectures, Razorpay payments, and instructor analytics. Nestor AI assistant for navigation and programming help, real-time Codespace with Monaco editor, community chat forums, and Cloudinary chunked video streaming with piracy protection. React + Zustand frontend, Node.js + Express backend with MongoDB, Socket.IO, and Gemini.

React Zustand Socket.IO MongoDB Gemini Razorpay Cloudinary Node.js

FAQ Nobody Asked

Honest answers about how I work, how I got here, and what I do when Docker is behaving. No synergize allowed.

I just kept getting curious. Started with CSS... then frontend, backend, AI, FastAPI, Docker. Every time I learned something, it introduced me to five more things.

I don't really chase technologies. I just work on whatever helps solve the problem. The stack changes on its own.

I don't actually hate Java. My college professor just made sure our relationship had a terrible start 😂

I can go from watching Kubernetes videos to "Why does this country have two capitals?" in under 10 minutes.

Mostly infrastructure. I reached the point where I'm more interested in how software runs than how it looks.