Agentic AI Engineer Job in Hyderabad (Hybrid)

HinduTone is hiring. This is a current opening at HinduTone, posted on 17 October 2026.

Job details

Employer: HinduTone
Role: Agentic AI Engineer
Location: Hyderabad, Telangana, India
Work mode: Hybrid
Job type: Full-time
Experience: Experience building LLM agents or tool-using AI workflows
Last date: 30 November 2026
Apply: Call / WhatsApp +91-9666019191 or the HinduTone contact page

About the role

A lot of the work behind HinduTone is repetitive but careful: checking festival dates against sources, preparing temple page drafts, updating calendar entries, tagging careers articles and moving content between English, Hindi and Telugu. Much of this can be handled by AI agents that call tools, follow steps and ask a person before doing anything important. We are hiring an Agentic AI Engineer to build those agents safely and reliably.

This is a full-time, hybrid position based in Hyderabad. You will design multi-step workflows, connect models to internal tools, and make sure every agent has limits, logs and a clear point where a human approves its work. An agent that silently publishes a wrong date is worse than no agent at all, so guardrails and failure handling are as much a part of the job as the clever parts.

The role suits engineers who have already built tool-using LLM workflows, who think in terms of what can go wrong at each step, and who like making automation that editors actually trust.

What you will do

Design and build tool-using agents that help editors research, draft, check and update content across the website's festival, temple, panchang, careers and community sections.
Define tool interfaces with clear inputs, outputs and error messages so that models call them correctly.
Build and maintain MCP (Model Context Protocol) servers that expose internal data and actions to agents in a controlled way.
Orchestrate multi-step workflows, including branching, retries, timeouts and resumption after a failure.
Add guardrails such as allowed-action lists, input validation, rate limits and spending caps.
Build human-in-the-loop approval steps so that nothing is published or changed on the live site without a person signing off.
Instrument agents with traces and logs so every step, tool call and decision can be reviewed later.
Handle failures gracefully: detect loops, stop runaway runs, and report clearly what was and was not completed.
Evaluate agent runs against test scenarios and track success rates before and after changes.
Work with editors to understand which tasks are worth automating and which should stay manual.

Skills we are looking for

Must have

Experience building LLM agents or tool-using AI workflows that ran on real tasks.
Strong Python skills, including asynchronous code and working with APIs.
Practical understanding of tool or function calling and how models choose between tools.
Experience designing multi-step workflows with state, retries and error handling.
A security-minded approach to permissions, secrets and what an agent is allowed to touch.
Comfort reading logs and traces to work out why an agent behaved the way it did.

Good to have

Experience writing or running MCP servers.
Familiarity with workflow or orchestration frameworks, or experience building a simple one yourself.
Knowledge of observability tools and structured logging.
Experience designing approval queues or review interfaces for non-technical users.
Exposure to evaluation methods for agents, such as scenario suites and task success scoring.
Basic knowledge of content management systems and publishing workflows.

Who should apply

We are looking for engineers who have built LLM agents or tool-using AI workflows, whether at work or in substantial personal projects that ran on real tasks. A degree in computer science, IT or a related field is useful, but equivalent practical skills demonstrated through working systems are just as welcome. If you are new to agents and still learning the basics of LLM APIs, our fresher-level listings may suit you better. You should be able to come to our Hyderabad office on planned days and work from home on the rest.

How you will grow

You will shape how HinduTone uses automation, from the first tool design to the rules that decide when an agent must stop and ask. The work spans research, editing, data updates and publishing, so you will build agents for very different kinds of tasks rather than one narrow use case. You will also own the safety model for these agents, which means your judgement on permissions and approvals will directly affect how far the team can rely on automation. As the agents prove themselves, the range of workflows you design will widen.

Working side by side with editors also teaches something that is hard to learn from tutorials: how real people decide whether to trust an automated helper. You will see which logs they read, which approvals feel like a burden and which feel reassuring, and you will refine your designs accordingly. That understanding of trust, alongside the technical depth, is what separates agents that get used from agents that get switched off.

Preparing your application

Describe one agent or tool-using workflow you built, the tools it could call and the task it completed.
Explain how you handled failure in that system: loops, bad tool output, timeouts or partial completion.
Share code on GitHub or a technical write-up showing tool definitions and orchestration logic.
If you have built an MCP server or similar integration layer, describe what it exposed and how you restricted access.
Give an example of a human approval step you designed and why you placed it where you did.
Include any success-rate or evaluation numbers you tracked for your agents.

Before the interview

You may find our AI engineer jobs explainer helpful for revisiting the wider skill set that agent roles build on. Spend some time refreshing tool calling patterns, workflow state handling and common guardrail designs, and be ready to sketch how you would stop an agent that keeps repeating the same failed step.

How to apply

Share your updated resume with your relevant experience, key skills and links to work you can show, such as GitHub repositories, deployed projects or technical write-ups.

Call / WhatsApp: +91-9666019191
Online: use the HinduTone contact page, choose “Other” as the reason and attach your resume (PDF, up to 5 MB)
Subject line: Application for Agentic AI Engineer – tool-using agents

HinduTone never charges a fee for applications, interviews or joining. If anyone asks you for money in HinduTone’s name, do not pay and report it through our contact page.

Frequently asked questions

Can freshers or students apply for the Agentic AI Engineer role?

We are looking for people who have already built LLM agents or tool-using AI workflows. Freshers and students who are still learning LLM basics will find our fresher-level AI listings a better match for now.

Will agents publish content on their own?

No. Every agent you build will include human-in-the-loop approval, so editors review and sign off before anything is published or changed on the live site. Guardrails, logs and failure handling are core parts of the role.

What is the salary for this Agentic AI Engineer job?

The salary is not published. It is discussed with shortlisted candidates after the first conversation, based on your skills and the experience you bring to the role at HinduTone.

Is this role hybrid, and where is it based?

Yes. The role is hybrid and based in Hyderabad, Telangana. You will work from the HinduTone office in Hyderabad on some days and from home on others, so you should be able to travel to Hyderabad regularly.

What is the last date to apply?

The last date to apply is 30 November 2026. We review applications as they arrive, so sending yours earlier gives us more time to read your work and get back to you.

How do I apply for this Agentic AI Engineer job?

Call or WhatsApp +91-9666019191, or send your resume through the HinduTone contact page with the reason set to Other, using the subject line 'Application for Agentic AI Engineer – tool-using agents'. HinduTone never charges a fee at any stage.

Editor's note (17 October 2026)

This is an official HinduTone job opening. Salary is discussed with shortlisted candidates and is not listed here. If the role is filled or closed, this page will be updated.