Software Engineer

Abhinav Joshi

I build the load-bearing parts — auth, task queues, the service three other services quietly depend on.

Abhinav Joshi
1.5 yrShipping in production

About

I'm a backend-leaning software engineer who ends up owning whatever's load-bearing — auth, task queues, the service that three other services quietly depend on. Most recently that's meant pulling authentication out of a monolith, replacing a task queue under production LLM traffic without anyone noticing, and leading a frontend for a legal SaaS product I didn't start on.

I like systems with clear boundaries and code that's still legible after the person who wrote it has moved on — usually me, six months later.

Languages

JavaScript / TypeScript · Python · C# · C++

Development

React · Next.js · Express · Flask · FastAPI · ASP.NET Core

Infra/DevOps

AWS (EC2, EBS, RDS, ElastiCache) · Azure · Docker · Nginx · Kong · Keycloak

Data

MongoDB · PostgreSQL · Redis · NATS JetStream

Experience

Panaroma Intelligence Solutions

Software EngineerFeb 2025 — Present
  • —Migrated authentication out of the monolith backend into a standalone Python/FastAPI Auth Microservice with a self-hosted Keycloak identity provider, onboarding 12 organizations (58 users) across 3 products in production with multi-tenant, organization-level access segregation and Casbin-based RBAC.
  • —Re-architected the core backend's distributed task processing pipeline, migrating from Celery/Redis to Taskiq running on NATS JetStream, preventing duplicate execution of long-running LLM calls with an early-acknowledgment, durable pull-based consumer model split across three dedicated task queues.
  • —Owned an analytics microservice serving 2 products off a shared FastAPI router, using a per-request resolver and DI container to dispatch each route to product-specific services behind a common service protocol.

InvoLead Services Pvt. Ltd.

Software Developer InternMay 2024 — Jul 2024
  • —Contributed to rebuilding a React product in TypeScript and Next.js 14.
  • —Used Next.js server-side capabilities to write end-to-end authentication.
  • —Used ChromaDB and Node.js for data embedding, vector storage, and retrieval in the product's RAG pipeline.

Projects

KubeCentrix: Fraud Detector thumbnail
JavaScriptManifest v3FlaskMachine Learning

KubeCentrix: Fraud Detector

Hackathon-winning Chrome extension (Rajasthan Police Hackathon 1.0) that blocks fraudulent URLs using SSL checks, keyword analysis, and an ML model trained to 94.6% accuracy, backed by a Flask REST API.

GitHub
Reactflow-X-DAG thumbnail
JavaScriptReactZustandReactFlowPythonFastAPI

Reactflow-X-DAG

A visual pipeline builder using React Flow and Zustand for the canvas and state, with cycle detection and DAG analysis handled by a FastAPI backend.

GitHub
FinCentrix: AI Powered Financial Advisory thumbnail
JavaScriptTypeScriptNext.jsExpressChromaDBDockerMicrosoft Azure

FinCentrix: AI Powered Financial Advisory

FinCentrix is an innovative platform that offers personalized financial guidance, transforming how people handle their finances.

GitHub
Google Search Automation thumbnail
PythonTkinterGUI

Google Search Automation

A simple Python software to automate Google searches.

GitHub
RAG-X-Langchain-JS thumbnail
JavaScriptNode.jsChromaDBLangchainJS

RAG-X-Langchain-JS

RAGXLangchainJS is a Multi-Doc RAG Application made using JavaScript, Node.js, LangChainJS, and ChromaDB to make your life easier.

GitHub

Writing

All posts →