Skip to content

Available for AI & backend roles

I build production-gradeagentic systems and theinfrastructure they run on.

AI Systems & Backend Engineer. I work on multi-agent LLM orchestration, real-time and event-driven backends, and money-movement infrastructure — architecting, shipping, and operating them end to end.

View selected work
Arnav Kumar
Arnav KumarIndia/—
AI conversations / month
0+
Monthly active users
0+
Production platforms
0

01About

I turn hard systems problems into reliable products.

My signature across every system: event-driven backends, strict multi-tenant and session isolation, governance and whitelist patterns, and observability baked in from day one.

AI Systems & Backend Engineer specializing in production agentic systems, multi-agent LLM orchestration, real-time workflows, and event-driven infrastructure. I've architected an AI financial advisory platform, a telephony + billing operations platform, and scalable backend services — owning architecture, observability, reliability, and deployment end to end.

My deepest work is agentic LLM orchestration. My widest-surface work is a telephony, billing, and attribution platform I architected and built solo — routing calls, moving money, and reconciling both.

Reach for most

Google ADKFastAPINode.jsGoPostgreSQLGCP
  1. 01

    Agentic AI systems

    Multi-agent orchestration on Google ADK and Gemini — coordinators, specialists, tool calling, deterministic grounding, and per-agent LLM observability.

  2. 02

    Real-time & event-driven backends

    Socket.IO call routing, webhook pipelines, Bull and Redis job systems, and infrastructure that stays correct under concurrency.

  3. 03

    Billing & money-movement infrastructure

    Multi-tenant Stripe Connect billing, payouts, reconciliation, and server-trusted pricing engines built to be auditable.

  4. 04

    End-to-end ownership

    From architecture and data modelling through observability, reliability, and deployment — I ship and then operate what I build.

02Experience

Where I've shipped.

Professional roles and engagements. Personal and contract project case studies live in the next section.

  1. Jan 2026 — Present

    Full-time

    AI Systems Engineer

    Sure Financial

    Architected a production AI financial advisory platform and the engagement infrastructure around it.

    • Architected a production AI financial advisory platform using Google ADK, Gemini, tool-calling, session memory, and real financial-data grounding.
    • Built a coordinator → specialist multi-agent system (routing, conversational advisor, tool/UI actions, and a loan-negotiation journey) with a parallel 'thinking' agent for fast perceived responses.
    • Eliminated async context leakage across concurrent user sessions using Python ContextVar isolation.
    • Designed structured LLM output schemas (version-gated by client build) enabling dynamic Flutter widget rendering on mobile.
    • Reduced hallucinations via deterministic data grounding — compact context cards plus a post-response check that verifies any institution named actually exists in the user's accounts.
    • Implemented human-in-the-loop flows (ADK long-running tools) and real-time escalation webhooks for expert consultation.
    • Integrated Langfuse observability with per-agent cost attribution and routing-drift detection.
    • Built FastAPI services and orchestration infrastructure supporting 1,000+ monthly active users and 4,000+ AI conversations per month.
    • Built a dedicated Go (Gin/GORM) gateway microservice to MoEngage for user-profile and event sync, with an internal event-governance catalog enforcing a single, consistent event vocabulary.
    Google ADKGemini 2.5 Pro/FlashTool CallingLangfuseDeterministic RAGFastAPIPythonasyncioGoGinGORMPostgreSQLSQLAlchemyFirebase FirestoreCloud SQLGCP / Cloud RunMoEngage

  2. 2025

    Contract

    Lead Backend Engineer

    inboundSelect (US client)

    Solely architected and built a production inbound-call operations platform — routing, billing, attribution, and marketplace workflows.

    • Solely architected and built a multi-tenant inbound-call operations platform handling routing, billing, payouts, attribution, and marketplace workflows.
    • Engineered low-latency call routing using Twilio, Socket.IO, and an eligibility-based dispatch engine.
    • Developed multi-tenant billing and payout systems using Stripe Connect with financial isolation per agency.
    Node.jsExpressSocket.IOTwilioStripe ConnectBullRedisPostgreSQLAWS SESGoogle Ads APIRailwayVercel

    Contract engagement. A deeper technical case study appears in Projects.

  3. Apr 2025 — Jun 2025

    Internship

    Backend Engineer Intern

    Exploria

    Built scalable REST backends and AWS-integrated workflows.

    • Designed and implemented REST APIs using Node.js and Express.
    • Integrated AWS S3 and Lambda into scalable backend workflows.
    • Built API validation, testing, debugging, and documentation processes.
    Node.jsExpressRESTAWS S3AWS Lambda

03Selected work

Systems I built end to end.

Each of these is a real production system. Open one for the full case study — the problem, the architecture, and the decisions behind it.

  1. 012025

    In production

    InboundSelect

    Multi-tenant pay-per-call operations platform

    Real-time inbound-call routing, multi-tenant Stripe Connect billing & payouts, marketplace resale, and Google-Ads attribution — built end to end.

    Tenancy
    Multi-tenant
    Billing tiers
    3
    DB migrations
    100+
    Routing
    Real-time
    Node.jsExpress 5Socket.IOTwilioStripe ConnectBullRedis (ioredis)

  2. 022025

    Live

    Royal Algae

    D2C e-commerce platform for premium sea-moss wellness

    A complete D2C storefront — guest phone-OTP checkout, Razorpay payments, server-side pricing, invoicing, and an admin dashboard.

    Checkout
    Guest + auth
    Payments
    Razorpay + COD
    Pricing
    Server-trusted
    Stack
    Laravel + React
    Laravel 12PHP 8.2SanctumSpatie PermissionRazorpayMsg91 SMSAWS S3

  3. 032025–2026

    In production

    BuildFlowCRM

    Commercial real-estate CRM with maker-checker governance

    A MERN CRM for commercial real estate — five asset classes modeled polymorphically, plus an approval workflow where every employee edit and delete is reviewed before it lands.

    Asset classes
    5
    Approvals
    Maker-checker
    Data I/O
    Excel round-trip
    Stack
    MERN
    Node.jsExpressMongoose 8 (discriminators)JWT (httpOnly cookies)helmetexpress-rate-limitCloudflare R2 (S3 API)

04Toolkit

What I reach for.

Grouped by where it sits in a system rather than by how well I know it — everything listed here is something I've shipped with.

AI & Agent Systems

Google ADKGeminiMulti-Agent OrchestrationTool CallingRAGLangGraphLangfusePrompt Engineering

Backend & Systems

Node.jsExpressFastAPIGo (Gin)REST APIsWebSockets / Socket.IOBull QueuesEvent-Driven SystemsReal-Time InfrastructureJWT Auth

Languages

JavaScript / TypeScriptPythonGoPHPJavaSQL

Databases & Storage

PostgreSQLMongoDBRedisSupabaseFirebase

Cloud & Infra

AWS (S3, SES, Lambda)GCP / Cloud RunVercelRailwayGitHub ActionsCI/CDCloudflare

Integrations & Tooling

TwilioStripe ConnectRazorpayGoogle Ads APIGoogle Calendar APIMoEngagePostmanGit

05Open source

Published & shared.

Tools I've released for other developers to use.

06Contact

Let's build something.

Hiring, a contract build, or you just want to talk systems? Send a message — or ask my AI assistant first and it'll pass your details along.

Email
arnav@arnavlogic.com
Based in
India
Status
Open to work
Elsewhere
GitHubLinkedIn

Open to

  • AI / agentic systems engineering roles
  • Backend / platform engineering roles
  • Contract & freelance builds (LLM apps, real-time systems, billing infra)
  • Technical collaborations