Data & Artificial Intelligence3 - 6 Years
AI & Machine Learning Engineer
Join our AI innovation team to train, fine-tune, and deploy large language models (LLMs), semantic search embeddings, and predictive intelligence systems. You will bridge advanced ML research and high-throughput production deployment.
Salary Benchmark
₹18,00,000 - ₹40,00,000 LPA ($120,000 - $185,000 USD)
Experience Range
3 - 6 Years
Key Roles & Responsibilities
- Fine-tune generative AI models (Gemini, Llama 3, Claude) using LoRA/QLoRA techniques.
- Develop high-performance Retrieval-Augmented Generation (RAG) pipelines with Vector databases (Pinecone, pgvector).
- Deploy low-latency model inference servers with vLLM, Triton, or ONNX Runtime.
- Monitor model drift, hallucination metrics, and response latency in production.
- Collaborate with data engineering to clean, curate, and vectorize massive datasets.
Must-Have Requirements
- ✓Bachelor's or Master's degree in Computer Science, Data Science, or related quantitative field.
- ✓Proficiency in Python, PyTorch, LangChain, LlamaIndex, and Hugging Face Transformers.
- ✓Hands-on experience with vector embeddings, cosine similarity, and hybrid BM25 search.
- ✓Familiarity with containerized model serving on GPU-accelerated cloud nodes (AWS EC2 G5/P4).
Nice-to-Have Skills
- +Published research in NLP, LLM alignment, or Agentic workflows.
- +Experience deploying custom quantized models on edge or private LAN hardware.
Recruiter Interview Guide
Phone Screening Questions & Evaluation Rubric
Use these questions to vet candidates during initial phone screens before submitting shortlists to clients.
Q1. How do you prevent and detect hallucinations in an enterprise RAG system?
Good Candidate Signal:
Mentions ground truth verification, Ragas metrics, temperature settings, and reranking top chunks with cross-encoders.
Red Flag:
Believes prompting alone completely eliminates model hallucinations.
Ready to Source Candidates for this Role?
Post this job mandate into your Ascent ATS workspace, auto-parse applicants with Gemini AI, and generate client magic shortlist links in minutes.