AI Engineer Job Description

About the Role

We are looking for a skilled AI Engineer with a strong focus on Generative AI to design, build, and ship production-grade AI systems. You will work at the intersection of large language models (LLMs), multi-modal AI, and software engineering taking GenAI capabilities from prototype to reliable, scalable products used by real users. You will collaborate closely with product, data, and platform teams and have end-to-end ownership of GenAI features and infrastructure.

Key Responsibilities

Generative AI Development

- Design and build LLM-powered applications including conversational AI, content generation, summarisation, classification, and code generation systems.
- Engineer advanced prompt pipelines system prompts, few-shot examples, chain-of-thought, structured output, and tool-use patterns optimised for reliability and cost.
- Build and maintain Retrieval-Augmented Generation (RAG) architectures:
- Chunking strategies, embedding model selection, and semantic search tuning.
- Hybrid search (dense + sparse), metadata filtering, and re-ranking pipelines.
- Vector store management using Pinecone, Weaviate, Chroma, pgvector, or Qdrant.
- Develop and deploy AI agents and multi-agent systems:
- Tool-calling agents with function use / ReAct patterns.
- Long-horizon planning agents with memory and reflection loops.
- Orchestration using LangChain, LlamaIndex, AutoGen, or custom frameworks.
- Fine-tune and adapt foundation models for domain-specific tasks:
- Parameter-efficient fine-tuning (LoRA, QLoRA, adapter layers).
- Instruction tuning, supervised fine-tuning (SFT), and preference alignment (RLHF, DPO, PPO).
- Dataset curation, annotation pipeline design, and data quality validation.
- Work with multi-modal models including vision-language models (VLMs), image generation, and speech/audio models as required by product needs.
- Evaluate and benchmark GenAI outputs using systematic evaluation frameworks:
- Automated metrics (ROUGE, BERTScore, G-Eval, LLM-as-judge patterns).
- Human evaluation rubric design and inter-annotator agreement analysis.
- Red-teaming, adversarial probing, and safety/alignment testing.

ML Engineering & Deployment

- Package, deploy, and serve AI models at scale using cloud-native tooling (SageMaker, Vertex AI, Azure ML, or equivalent).
- Manage inference infrastructure — optimise latency, throughput,



and cost using quantisation, batching, caching, and model distillation.
- Build and maintain data and feature pipelines for training, evaluation, and real-time inference.
- Implement observability for AI systems: token usage tracking, latency monitoring, hallucination detection, and output quality dashboards.
- Integrate third-party LLM APIs (OpenAI, Anthropic, Google Gemini, Cohere, Mistral) and manage prompt versioning and fallback strategies.

Collaboration & Communication

- Partner with product managers and designers to translate user requirements into GenAI feature specifications.
- Communicate model trade-offs, risks, and evaluation results clearly to technical and non-technical stakeholders.
- Contribute to internal knowledge sharing through documentation, demos, and technical write-ups.

Generative AI — Technical Skills

Must-Have (Required)

- LLM APIs & SDKs: Proficient in OpenAI API, Anthropic Claude API, or equivalent; experience with streaming, function calling, and structured output modes.
- Prompt Engineering: Advanced techniques including chain-of-thought, self-consistency, few-shot prompting, output formatting, and prompt injection mitigation.
- RAG Systems: Hands-on experience building end-to-end RAG pipelines, including document ingestion, chunking, embedding, indexing, retrieval, and answer synthesis.
- Vector Databases: Practical experience with at least one (Pinecone, Weaviate, Chroma, pgvector, Qdrant).
- LLM Fine-Tuning: Familiarity with PEFT techniques (LoRA, QLoRA); experience fine-tuning on HuggingFace Transformers or equivalent.
- Agent Frameworks: Experience building tool-use or ReAct-style agents using LangChain, LlamaIndex, or custom orchestration.
- Evaluation: Ability to design and run systematic evals; familiarity with LLM-as-judge, G-Eval, or similar frameworks.
- Python: Solid Python skills with experience in async programming, API design, and production-quality code.

Nice to Have

- Preference alignment: RLHF, DPO, PPO,



or Constitutional AI methods.
- Multi-modal models: Vision-language models (GPT-4V, LLaVA, Gemini Vision), image generation (Stable Diffusion, DALL-E), or speech models (Whisper, ElevenLabs).
- Model optimisation: Quantisation (GPTQ, AWQ, bitsandbytes), speculative decoding, KV cache tuning.
- Serving infrastructure: vLLM, Triton Inference Server, TGI (Text Generation Inference), or Ollama for self-hosted models.
- Knowledge graphs and structured memory systems for long-context agents.
- Guardrails and safety tooling: NeMo Guardrails, LlamaGuard, or custom content moderation pipelines.
- Multi-agent orchestration: AutoGen, CrewAI, or custom multi-agent workflows.
- Embedding model fine-tuning for domain-specific retrieval quality.

General Engineering Skills

- 2–4 years of professional software or ML engineering experience.
- Cloud platform proficiency: AWS, GCP, or Azure — model hosting, storage, serverless functions, and managed ML services.
- API development: RESTful and/or GraphQL API design; experience integrating AI capabilities into backend services.
- Experiment tracking and versioning: Weights & Biases, MLflow, or DVC.
- Software engineering fundamentals: Git, CI/CD pipelines, unit and integration testing, code review.
- Containerisation basics: Docker for packaging and deploying ML services.
- SQL and data literacy: Comfortable writing queries and working with structured data sources.

What We Look For

- GenAI product intuition — understands not just how to build with LLMs but when to and when not to; can identify where GenAI genuinely adds value vs. adds complexity.
- Evaluation-first mindset — insists on measuring output quality before and after changes; treats evals as a core engineering discipline, not an afterthought.
- Curiosity and velocity — follows AI research closely, experiments rapidly, and has clear opinions on emerging techniques worth adopting.
- Shipping mindset — moves from prototype to production confidently; understands the gap between a demo and a reliable product.
- Responsible AI practice — proactively considers hallucination risks, bias, safety, and data privacy in every system they build.
- Clear communication — able to explain model behaviour, failure modes, and trade-offs to product and business stakeholders without jargon.

📌 Artificial Intelligence Engineer (Bengaluru)
🏢 Aptus Data LAbs
📍 Bengaluru