abhishek/runtimeresume
Role targetSDE / Backend / AI Systems
Current modeOpen to full-time roles
BaseNagpur, India
Builder biasShip from scratch, then harden
Abhishek Raut
operatoronline

Abhishek Raut

Backend. AI systems.
Deployable products.

inspect buildsstart a thread
system mapscratch to scale
APIFastAPI / Node
StatePostgres / Redis
AgentsLangGraph / MCP
ToolsRAG / workers
DeployDocker / AWS
ObservePrometheus / Grafana
5+Agentic AI systems shipped
60%Order processing time reduced
10msPlayback sync target reached
1500+DSA problems solved
build feed

$ bootstrap product idea

> model domain, data, workflows

> build API and persistence first

> wire agents only where they earn trust

> containerize, deploy, observe, iterate

ok ship useful systems, not demos

01
TypeScriptReactNext.jsTailwind CSSZustand
02
PythonFastAPINode.jsExpress.jsWebSockets
03
PostgreSQLMongoDBRedisPrismaChromaDB
04
DockerKubernetesAWSArgo CDGitHub Actions
05
LangGraphLangChainMCPRAGFine-tuning

// build queue

Systems that prove the builder mindset.

Projects are shown as engineering artifacts: constraints, architecture, and measurable behavior.
open full systems archive ->

SYNC-ROOM / 2025

PlaySync

source

A real-time music platform where users join rooms and keep playback synchronized across clients with low drift.

architecture

Turborepo with Next.js, Node.js, Socket.io, Redis pub/sub, PostgreSQL, Prisma, and NextAuth.

proof

Served 1,000+ registered users and reduced sync drift from 120ms to around 10ms at 5K+ messages/sec.

role

Designed the real-time architecture, data model, authentication, and Redis-backed synchronization path.

TypeScriptTurborepoNext.jsNode.jsRedisNextAuthPostgreSQLPrismaSocket.io

// build protocol

Engineering from the spine out.

01

model the domain

Data, workflows, and failure paths first.

Define ownership, edge cases, and system contracts before writing a single endpoint.

PostgreSQLPrismaevent modelingschema design
02

build the working spine

API, auth, persistence, cache, background work.

Ship the hardest parts early. Auth, data layer, job queues, and CI/CD path come before polish.

FastAPINode.jsExpressRedisJWTWebSockets
03

deploy and harden

Container on day one. Push to production without fear.

CI/CD, containerization, and infrastructure wired before feature work accelerates.

DockerKubernetesAWS ECSGitHub ActionsNginxArgo CD
04

measure and iterate

Numbers first, then optimization.

Trace sync drift, latency, re-renders, and throughput. Fix what data confirms.

PrometheusGrafanaload testingprofiling
architecture layerslive

INGRESS

Nginx · AWS ALB · Route 53

API LAYER

FastAPI · Node.js · Express · gRPC · WebSockets

STATE + CACHE

PostgreSQL · MongoDB · Redis · Prisma

AI AGENTS

LangGraph · MCP · ChromaDB · Qdrant · RAG

DEPLOY

Docker · Kubernetes · AWS ECS · Argo CD

OBSERVE

Prometheus · Grafana · GitHub Actions

// ai control plane

AI systems as workflows, tools, memory, and evaluation.

No magic wrapper energy. The agent earns its place by doing useful work inside a reliable system.
01Intent router
02LangGraph state machine
03MCP-compatible tools
04RAG / semantic memory
05LoRA fine-tuning
06Trace and evaluate
agent traceLangGraph / MCP / RAG
orchestrator

Routes user intent to domain agents and structured tools.

state store

Maintains multi-turn context across Redis-backed sessions.

retrieval

Pulls historical incidents, product data, or support knowledge into the loop.

tool execution

Runs MCP-compatible tools, API calls, and workflow actions with guardrails.

evaluation

Checks root-cause quality, response relevance, and human handoff reduction.

production log

Ignix Social / Software Engineer

Jan 2025 - Present
01

Led an Order Management System for a fashion designer and cut order processing time by 60%.

02

Built QR-based restaurant SaaS for contactless menu browsing and ordering across multiple clients.

03

Delivered LLM fine-tuning on customer support data to improve relevance and reduce human handoff.

04

Designed and shipped 5+ agentic AI systems using LangGraph, LangChain, and LoRA fine-tuned models.

05

Built ERP modules with TypeScript, React, and Zustand; cut unnecessary re-renders by 40%.

signal ledger

1879+LeetCode rating
Top 5%Competitive programming
Rank 13InterviewBit global
40+Engineers mentored

Competitive programming gave me the patience to reason through edge cases. Product work gave me the instinct to turn that reasoning into systems people can actually use.

// open channel

Bring me hard systems, early products, and useful AI problems.

email: Abhishekgajananraut@gmail.com

github: abhishekraut01

linkedin: sweabhishek