Skip intro

01 / 09

AI / Agent EngineerCloud-Native Systems

Shubham Patel

Forward Deployed Engineer

Deployed at Technotribes.aiGurugram, India4+ yrs

Engineering what comes next.

I build what happens behind the interface.

Impact in numbers

  • 4+ years in production, shipped, not demoed. Python and Java services — FastAPI, Django, Spring Boot — running for real users.
  • 30% cloud cost, removed. A rightsizing module that found over-provisioned EC2 and RDS instances.
  • 7s to tag resources, down from sixty. AWS and Azure provisioning automated with Ansible and the AWS SDK. Deployment time fell 20%.
  • 1,000+ people protected, in the browser. A React tool that masks Aadhaar numbers client-side, before anything reaches a server.
  1. Interface: React clients that respect the system behind them. React, JavaScript, Teams Adaptive Cards
  2. APIs & Identity: Typed contracts, with identity enforced at the door — SAML SSO, RBAC, JWT. FastAPI, Spring Boot, Entra ID · SAML 2.0
  3. Services: Small services behind a gateway, with discovery and central configuration. Java, Python, Microservices
  4. Data & Messaging: State, and how it moves — queues that absorb load and dead-letter what fails. PostgreSQL · MongoDB · Redis, RabbitMQ · Kafka · SQS
  5. Cloud: Provisioned by automation, orchestrated on Kubernetes, rebuilt the same way every time. AWS · EKS, Kubernetes (CKA), Docker · Ansible · CI/CD

Beyond request and response,

From software systems to intelligent systems.

Agentic orchestration graphUser Request → Intent Router → Supervisor → Specialist Agents → Tools · MCP → Validation → Response. A dashed clarification loop returns from the router to the request; the supervisor fans out to three specialist agents, each calling its own tool, and all tools converge on validation before the response.RequestRouterSupervisorRetrievalAnalysisActionRAG · SearchDataAPIsValidationResponse
  1. 01

    User Request

    A goal arrives in plain language — ambiguous, incomplete, human.

  2. 02

    Intent Router

    The request is classified. If something essential is missing, the system asks rather than guesses.

  3. 03

    Supervisor

    A supervising agent plans the work and decides which specialists to involve, and in what order.

  4. 04

    Specialist Agents

    Each agent has one bounded responsibility and its own context — small enough to trust.

    Retrieval · Analysis · Action

  5. 05

    Tools · MCP

    Agents reach real systems through typed tool calls: search, data, APIs — exposed over MCP.

    RAG · Search · Data · APIs

  6. 06

    Validation

    Output is checked against schema and intent before anything leaves the system.

  7. 07

    Response

    One coherent, grounded answer — with the work behind it accounted for.

  1. Development: Small commits, typed contracts, tests written beside the code.
  2. Quality assurance: Regressions, edge cases and load, checked before anyone else sees it.
  3. User acceptance: The people who will use it run their real workflows and sign off.
  4. Production: Containerised, deployed and watched once it is live.

Selected work

Selected work — 01 / 03

  • 04Agent stages
  • SSOEntra ID · OBO
  • MCPFile delivery

Agentic AI platform in Microsoft Teams

Requirements Automation

A multi-agent workflow that drafts business and functional requirements from a chat.

Technotribes.ai · 2026

Microsoft Foundry · Teams · Microsoft Graph · Azure AI Search · MCP · Python

Selected work — 02 / 03

  • 7sTagging, from 60s
  • 20%Faster deploys
  • SSOSAML 2.0 · RBAC

Financial SaaS platform

Finalyze.ai

Owned from design through production: identity, provisioning automation and cloud cost.

Ataloud Technologies · 2022 – 2025

Java · Spring Boot · React · PostgreSQL · MongoDB · RabbitMQ · Redis · AWS · Kubernetes

Selected work — 03 / 03

  • RBACRole-based clients
  • GWAPI gateway
  • SDService discovery

Services and APIs for a securitization product

Securitization

Spring Boot services behind a gateway, serving role-based React clients.

Ataloud Technologies

Java · Spring Boot · Spring Cloud · API Gateway · RBAC · React

  1. Requirements Automation

    Agentic AI platform in Microsoft Teams. A multi-agent workflow that drafts business and functional requirements from a chat. Technotribes.ai · 2026

    04 Agent stages, SSO Entra ID · OBO, MCP File delivery

    Microsoft Foundry, Teams, Microsoft Graph, Azure AI Search, MCP, Python

  2. Finalyze.ai

    Financial SaaS platform. Owned from design through production: identity, provisioning automation and cloud cost. Ataloud Technologies · 2022 – 2025

    7s Tagging, from 60s, 20% Faster deploys, SSO SAML 2.0 · RBAC

    Java, Spring Boot, React, PostgreSQL, MongoDB, RabbitMQ, Redis, AWS, Kubernetes

  3. Securitization

    Services and APIs for a securitization product. Spring Boot services behind a gateway, serving role-based React clients. Ataloud Technologies

    RBAC Role-based clients, GW API gateway, SD Service discovery

    Java, Spring Boot, Spring Cloud, API Gateway, RBAC, React

Expertise — 08

A stack chosen for the problem.

  • 01 / 06

    Orchestration, grounding, evaluation.

    • Multi-Agent Orchestration
    • LangGraph
    • LangChain
    • RAG
    • MCP
    • LLM Evaluation
    • Guardrails

Open to conversations

Let's buildwhat comesnext.

Forward Deployed Engineer at Technotribes.ai, Gurugram. Reach out about agentic AI, cloud platforms or distributed systems.

Shubham

Patel

Contact pass

SP-2026CKA · AWS Academy · B.Tech CSE