LLM and RAG 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: LLM and RAG Engineer
Location: Hyderabad, Telangana, India
Work mode: Hybrid
Job type: Full-time
Experience: Experience building retrieval-augmented generation (RAG) systems
Last date: 30 November 2026
Apply: Call / WhatsApp +91-9666019191 or the HinduTone contact page

About the role

HinduTone publishes a large and growing library: temple profiles, festival explainers, daily panchang pages, deity names, careers articles and community stories, in English, Hindi and Telugu. Finding the right passage in that library, and answering a reader's question from it without inventing details, is the job of a well-built retrieval-augmented generation (RAG) system. We are hiring an LLM and RAG Engineer to design, build and measure that system.

The position is full-time and hybrid, based in Hyderabad. You will decide how content is split, embedded, indexed and searched, how results are reranked, and how answers point back to the pages they came from. A reader asking about a temple's timings or the date of a festival deserves an answer grounded in our published pages, with a clear citation, or an honest "we do not have this".

This role suits an engineer who has already built a RAG system beyond a demo, who cares about retrieval quality numbers, and who finds hallucination reduction an interesting engineering problem rather than an afterthought.

What you will do

Design chunking strategies suited to different content types, such as long festival articles, short temple fact blocks and date-based panchang entries.
Choose, test and run embedding models, including ones that handle Hindi and Telugu text well.
Build and maintain vector indexes, for example with pgvector, alongside the content database.
Combine keyword search and vector search into hybrid retrieval, and tune how the two are weighted.
Add reranking to improve the order of retrieved passages before they reach the model.
Make every generated answer cite the source pages it used, and refuse politely when the sources do not support an answer.
Build retrieval evaluation sets with real reader-style questions and track recall, precision and answer faithfulness over time.
Find and reduce hallucinations, especially around dates, timings, locations and names.
Keep indexes fresh as editors publish, update or remove pages, so stale content does not appear in answers.
Monitor query latency and index size, and plan for the library continuing to grow.
Document the pipeline clearly so that other engineers and editors understand how answers are produced.

Skills we are looking for

Must have

Experience building a retrieval-augmented generation system used by real people, not only a prototype.
Solid Python skills and comfort working with databases.
Practical knowledge of embeddings, vector similarity search and at least one vector store.
Experience with chunking choices and an understanding of how they affect retrieval.
Familiarity with hybrid search (for example BM25 combined with dense retrieval) and reranking models.
Ability to design retrieval evaluations and read the results honestly.
Good SQL skills, ideally with PostgreSQL.

Good to have

Hands-on use of pgvector or a similar extension inside PostgreSQL.
Experience with multilingual or cross-lingual retrieval, especially for Indian languages.
Knowledge of citation and grounding techniques that link answer sentences to source passages.
Experience with incremental indexing and change detection for content that is edited often.
Exposure to query rewriting, metadata filtering or date-aware retrieval.
Interest in Hindu culture, temples and festivals, which makes judging answer quality easier.

Who should apply

We are looking for engineers with real experience building RAG systems. A degree in computer science, IT, data science or a related field helps, though equivalent practical skills shown through delivered projects are equally welcome. If your experience is mostly with general backend work and you are only now learning retrieval, you may find a closer match in our other technology listings. You should be ready to work from our Hyderabad office on some days and from home on others.

How you will grow

You will be the main owner of how HinduTone finds and grounds information, which is a central piece of how readers will interact with the site. The content is varied, multilingual and constantly changing, so you will work on problems such as date-sensitive retrieval and cross-script matching that are hard to meet elsewhere. You will set the evaluation standards others follow, and your measurements will guide decisions about models, indexes and infrastructure. As the library grows, the scale and design choices you make will become more significant.

You will also work closely with editors who know the content best. Their feedback on which answers feel right, and which quietly miss the point, will sharpen your sense of retrieval quality in a way that metrics alone cannot. Many engineers find that this mix of numbers and editorial judgement is what makes the work satisfying, and it is a skill set that stays useful whatever tools and models come next.

Preparing your application

Describe a RAG system you built: what content it covered, how big the corpus was and who used it.
Explain your chunking and embedding choices for that system and what you changed after testing.
Share retrieval metrics you tracked, such as recall at a given cut-off or answer faithfulness, with before and after numbers.
Include a GitHub repository or a technical write-up showing hybrid search or reranking in practice.
Mention any work with multilingual text, and how you checked retrieval quality across languages.
Give one example of a hallucination you found and the fix that removed it.

Before the interview

Our explainer on AI engineering jobs and skills is a useful refresher on the fundamentals interviewers often return to, such as evaluation design and model trade-offs. Before you meet us, revise embedding similarity measures, hybrid ranking methods and reranker behaviour, and think through how you would keep an index accurate when pages are edited every day.

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 LLM and RAG Engineer – retrieval systems

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 LLM and RAG Engineer role?

This position expects experience building retrieval-augmented generation systems that people actually used. Students and freshers who have only built small demos are encouraged to look at our fresher and learning-focused technology listings instead.

Which retrieval topics matter most for this role?

Chunking, embeddings, vector search with tools like pgvector, hybrid keyword and vector search, reranking, citations and retrieval evaluation. Reducing hallucinated dates, timings and names is a central part of the work.

What is the salary for this LLM and RAG 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 LLM and RAG 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 LLM and RAG Engineer – retrieval systems'. 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.