Senior AI Engineer

LLMs & Conversational AI

Experience

5–10 years

Employment Type / Department

Full time | Artificial Intelligence / Engineering

About the Role

We are looking for a highly skilled Senior AI Engineer with deep expertise in Large Language Models (LLMs), conversational AI, and end-to-end AI system development. The ideal candidate will have hands-on experience building, fine-tuning, evaluating, and deploying production-grade AI models and intelligent chatbot systems.

This role requires strong expertise across the AI engineering lifecycle, with particular emphasis on LLM fine-tuning, synthetic dataset generation, prompt engineering, conversational state management, and multi-turn dialogue systems.

The candidate will be responsible for building AI systems that go beyond simple question-answering to deliver context-aware, consistent, personalized, and reliable conversations. They should be comfortable working across model training, data pipelines, inference optimization, RAG architectures, evaluation frameworks, and production deployment.

We are looking for an engineer who can make informed architectural decisions, independently solve complex AI problems, and translate business requirements into scalable AI-powered products.

Key Responsibilities1. LLM Fine-Tuning & Model Optimization

· Design and optimize LLM fine-tuning pipelines using SFT, PEFT, LoRA, and QLoRA for domain-specific behavior.

· Select appropriate base models and troubleshoot training quality, convergence, GPU utilization, quantization, and inference performance.

· Establish repeatable training, evaluation, model versioning, and deployment workflows.

2. Synthetic Dataset Generation & Data Engineering

· Build scalable pipelines to generate realistic, high-quality multi-turn conversational datasets for training and evaluation.

· Implement dataset schemas, automated validation, filtering, deduplication, balancing, and leakage prevention.

· Evaluate synthetic data for hallucinations, contradictions, repetition, coverage, and downstream model impact.

3. Prompt Engineering & LLM Behavior Design

· Design and optimize system prompts, instruction hierarchies, few-shot examples, and structured outputs.

· Improve instruction adherence, consistency, contextual understanding, tool calling, and handling of ambiguous inputs.

· Use prompt evaluation and regression testing to determine when prompting, fine-tuning, retrieval, or application logic is appropriate.

4. Conversational AI, Chatbots & Dialogue Systems

· Build intelligent, context-aware chatbots that support coherent multi-turn conversations and adapt to user intent.

· Design dialogue flows for information gathering, clarification, follow-up, task completion, and escalation while avoiding repeated questions.

· Integrate LLMs with APIs, databases, and tools, with appropriate guardrails for hallucinations, safety, and out-of-scope requests.

5. Conversational State Management & Context Engineering

· Design state models, state machines, and orchestration logic to track known information, unresolved questions, user preferences, and conversation progress.

· Build context assembly, persistence, summarization, and session recovery mechanisms that maintain consistency across turns.

· Separate application state, user profile data,



retrieved knowledge, and transient context; use deterministic logic for critical workflow decisions.

6. Retrieval-Augmented Generation (RAG) & Knowledge Systems

· Build production RAG pipelines using embeddings, vector databases, hybrid search, reranking, and metadata filtering.

· Improve retrieval relevance, factual grounding, source attribution, and knowledge freshness.

· Determine when RAG, fine-tuning, prompting, or a hybrid approach best fits the use case.

7. AI Evaluation, Testing & Quality Assurance

· Build evaluation frameworks and benchmark datasets for LLMs, prompts, fine-tuned models, and multi-turn conversations.

· Measure accuracy, relevance, instruction adherence, safety, consistency, and conversational coherence using automated and human evaluation.

· Analyze failures and establish regression tests, quality thresholds, and release criteria.

8. Production AI Engineering & Deployment

· Develop scalable AI services and APIs, optimizing latency, throughput, concurrency, and operating cost.

· Implement model serving, streaming, asynchronous processing, caching, retries, monitoring, and graceful failure handling.

· Apply MLOps/LLMOps practices for reproducibility, versioning, deployment, privacy, and secure data handling.

9. Technical Leadership & Collaboration

· Own technical design and delivery of AI initiatives from experimentation through production.

· Make architectural decisions, document trade-offs, mentor engineers, and establish engineering best practices.

· Collaborate with product, backend, and domain teams to translate business requirements into measurable AI outcomes.

Required Skills & Qualifications

· 5–10 years of professional experience in AI/ML engineering, NLP, applied machine learning, or related software engineering roles.

· Strong programming skills in Python, including production-quality code, testing, debugging, and performance optimization.

· Deep practical understanding of transformer architectures, attention mechanisms, tokenization, embeddings, context windows, and LLM behavior.

· Hands-on experience fine-tuning LLMs for domain-specific tasks.

· Practical experience with PEFT techniques, particularly LoRA and QLoRA, and frameworks such as Hugging Face Transformers, TRL, PEFT, or equivalent tools.

· Experience designing synthetic data generation pipelines and preparing high-quality conversational datasets.

· Strong prompt engineering skills, including system prompt design, structured outputs, and instruction-following optimization.

· Experience building multi-turn chatbots or conversational agents for real-world applications.

· Strong understanding of conversation state, context management, memory, dialogue orchestration, and session persistence.

· Experience with RAG architectures, embedding models, vector search, and knowledge retrieval.

· Familiarity with LLM evaluation, benchmarking, automated testing,



and model quality analysis.

· Experience building APIs and integrating AI capabilities into production applications.

· Understanding of model serving, inference optimization, GPU utilization, and deployment considerations.

· Strong problem-solving skills and the ability to independently investigate complex model and system failures.

Preferred Skills

· Experience working with open-source models such as Llama, Qwen, Mistral, Gemma, or equivalent.

· Experience with training and inference optimization on NVIDIA GPUs and CUDA-based environments.

· Familiarity with Unsloth, Axolotl, DeepSpeed, FSDP, vLLM, or similar training and serving tools.

· Experience with agentic workflows, tool calling, function execution, and multi-step orchestration.

· Familiarity with LangGraph, LangChain, LlamaIndex, or custom orchestration frameworks.

· Experience with evaluation frameworks such as DeepEval, Ragas, or custom LLM evaluation pipelines.

· Experience with experiment tracking and model lifecycle tools such as MLflow or Weights & Biases.

· Knowledge of Docker, CI/CD, cloud infrastructure, and deployment of GPU-backed services.

· Experience with PostgreSQL, Redis, asynchronous processing, and distributed application architectures.

· Experience in healthcare AI, clinical conversational systems, or other domains requiring high reliability, privacy, and domain-specific safeguards is an advantage.

What Success Looks Like

· Model quality: Better instruction following, domain adaptation, response consistency, and task completion.

· Dataset quality: Higher-quality synthetic training data with measurable improvements in downstream model performance.

· Conversational intelligence: More coherent multi-turn interactions, fewer repeated questions, and better handling of ambiguous inputs.

· State reliability: Accurate tracking of known information, conversation progress, and unresolved tasks across turns.

· Production readiness: Reliable inference, acceptable latency, controlled operating costs, and robust failure handling.

· Evaluation maturity: Repeatable benchmarks, regression testing, and data-driven model improvement.

· Engineering ownership: Independent delivery of end-to-end AI solutions with explicit technical documentation and measurable results.

Ideal Candidate Profile

The ideal candidate is not limited to prompt writing or API integration. They understand how to build the complete AI system—from data generation and model training to dialogue orchestration, state management, evaluation, and production deployment.

They can determine whether a problem is best solved through fine-tuning, better training data, prompt changes, retrieval, explicit state management, or a combination of these approaches.

They are comfortable experimenting with models, analyzing failures, building custom tooling, and translating research and experimentation into reliable product capabilities.

The strongest candidates will have demonstrable, hands-on experience improving a conversational AI system through fine-tuning, high-quality synthetic data, rigorous evaluation, and robust multi-turn state management.

Pay: ₹300,000.00 - ₹350,000.00 per month

Perks:

- Cell phone reimbursement
- Paid sick time

Work Location: In person

📌 Senior AI Engineer (Chandigarh)
🏢 Bizetikk 24 Private
📍 Chandigarh