TL;DR
Our client, a technology company building LLM-powered products, is hiring an AI Engineer in Bengaluru. You build LLM features end to end: retrieval-augmented generation pipelines, evaluation harnesses, fine-tuning where it earns its place, and the work that keeps latency and cost within budget. It is a software engineering role with models at its centre. The role is hybrid, Bengaluru.
The client is confidential at this stage. We will tell you who it is before your profile goes anywhere.
Key points
- Builds and ships LLM features used by customers, not research prototypes.
- Owns RAG pipelines, evaluation harnesses and, where justified, fine-tuning.
- Latency, reliability and inference cost are part of the job, not someone else's problem.
- Strong software engineering is the foundation; prompting alone is not enough.
- Hybrid, Bengaluru. Confidential client, named to you before any submission.
What the work is
An AI Engineer here takes a product requirement and delivers an LLM feature that is accurate, fast and affordable enough to run at scale. You own it from design through production monitoring.
The work covers five areas:
- LLM features. Designing prompts, structured outputs, tool calls and fallbacks, and integrating them cleanly into backend services and APIs. Models from OpenAI, Anthropic and Google, and open-weight models such as Llama or Qwen, are all in scope depending on the task.
- RAG pipelines. Ingestion, parsing, chunking, embedding and indexing of documents; hybrid search combining vector and keyword retrieval; re-ranking; and keeping indexes fresh and permission-aware. Tooling includes vector stores such as pgvector, Qdrant or Elasticsearch.
- Evaluation harnesses. Golden datasets, automated scoring with deterministic checks and model-graded rubrics, regression tests in CI, and production feedback loops. The client treats evaluation as the main engineering discipline, because without it nobody can tell an improvement from a lucky run.
- Fine-tuning where needed. Supervised fine-tuning or parameter-efficient methods such as LoRA when prompting and retrieval are not enough, for example to cut cost on a high-volume task or enforce a strict output format. Knowing when not to fine-tune matters as much.
- Latency and cost. Prompt and context trimming, caching, streaming, batching, model routing between large and small models, and serving optimisation with engines such as vLLM. You track cost per request and treat it as a product metric.
What the client is looking for
Engineers who have put LLM features in front of real users and can explain how they measured quality. The best signal is someone who describes a feature that underperformed, how the evaluation showed it, and what they changed.
Must-haves
- Strong software engineering in Python, with production services you have built and run.
- Hands-on experience shipping LLM features, including retrieval and structured outputs.
- A habit of building evaluation before claiming improvement.
- Working knowledge of APIs, databases and cloud deployment on AWS, Azure or Google Cloud.
Nice-to-haves
- Fine-tuning experience with Hugging Face tooling or similar.
- Experience with agentic patterns and frameworks such as LangGraph.
- Model serving and GPU inference optimisation.
- A background in search, recommendation or classic ML.
Typical experience is usually 3 to 7 years in software or ML engineering, with recent hands-on work on LLM applications.
How this role differs from adjacent roles
AI job titles overlap heavily. This is how this seat compares.
| Role | Main focus | Typical output |
|---|---|---|
| AI Engineer (this role) | LLM features, RAG, evaluation, cost | Production features and pipelines |
| ML Engineer | Training and serving classic or deep learning models | Model pipelines and infrastructure |
| Data Scientist | Analysis, experimentation, statistical models | Insights and models for decisions |
| Agentic AI Engineer | Multi-step agent systems and tool use | Agent platforms and orchestration |
| Research Scientist | New methods and models | Papers and prototypes |
Process
A practical process focused on how you build and evaluate.
- A conversation with SYTalent about the role and your experience.
- A technical conversation with the client's engineering lead.
- A practical exercise: building or improving a small RAG feature, with an evaluation that shows whether it got better.
- A system design discussion and team conversation, then offer.
Why this role is worth a look
You will work on LLM features that customers actually use, with real traffic, real cost constraints and a team that takes evaluation seriously. The role builds the skills the market is shortest of: shipping reliable AI, not demos. It is also a direct path towards agentic systems and AI architecture as your experience grows.
Related reading
Our AI, deep tech and quantum practice covers how these searches run, and agentic AI in backtesting shows why evaluation discipline matters when models make decisions. The Agentic AI Engineers and Head of Artificial Intelligence searches are related, and AI jobs in India maps the wider market.
Common questions
What does an AI Engineer do?
An AI Engineer builds software features powered by large language models and runs them in production. The work includes prompt and output design, retrieval-augmented generation pipelines, evaluation harnesses, occasional fine-tuning, and optimising latency and cost. It sits between backend engineering and machine learning, with strong software skills as the foundation.
What is the difference between an AI Engineer and an ML Engineer?
An ML Engineer usually trains and serves models, building data and training pipelines. An AI Engineer usually builds applications on top of pre-trained foundation models, focusing on retrieval, prompting, evaluation and integration, with fine-tuning used selectively. The skills overlap, and many engineers move between the two roles.
Do I need fine-tuning experience for this role?
No, it is a nice-to-have. The client expects most problems to be solved with good retrieval, prompting and evaluation, and uses fine-tuning only where it clearly improves cost or quality. Showing that you understand when fine-tuning is and is not worth it matters more than having done it.
Is this AI Engineer role hybrid or remote?
The role is hybrid, Bengaluru. You will spend part of the week with the team in the office and part working remotely. We will share the exact pattern once we have spoken with you and named the client. Because of application volume, only shortlisted candidates are contacted.
Interested in this role?
Apply once, to us. We confirm the client with you before anything is submitted. Because of application volume, only shortlisted candidates are contacted.