Key Responsibilities:

• Design, implement, and evaluate ML/DL models using PyTorch, TensorFlow, or similar frameworks
• Build and optimize LLM-based systems, including prompt-tuning, fine-tuning, and adapter-based training (e.g., LoRA, QLoRA)
• Can develop robust and scalable RAG pipelines. In-depth knowledge of embeddings and can work with vector databases like FAISS, Pinecone, Weaviate, etc.
• Construct and maintain Agentic AI workflows involving multi-step reasoning, tool calling, memory components, and planning logic.
• Work with Proprietary APIs, as well as open-source libraries and models.
• Develop modular and clean Python code, adhering to software engineering best practices (OOP, reusable components, testing).
• Implement scalable solutions in cloud environments (like AWS), leveraging GPU/TPU resources effectively.
• Design inference pipelines that are robust and optimized for latency and throughput.
• Collaborate with research and product teams to translate ideas into production-grade ML features.

Required Skills:

• 4+ years of experience in machine learning and deep learning, including building models from scratch.
• Has a track record of shipping ML solutions that scale in production.
• Strong proficiency in Python and deep understanding of software design principles.
• Proven experience with transformer-based architectures, LLMs, and embedding models.
• Hands-on experience with RAG systems, deep understanding of agent-based systems.
• Familiarity with LangChain, LlamaIndex, or similar frameworks.
• Experience with cloud platforms (AWS/GCP/Azure) and understanding of scalability, resource optimization, and model deployment.
• Familiarity with performance profiling, efficient model serving, and hardware-aware design (e.g., GPU utilization, quantization).
• Ability to read, debug, and contribute to complex ML/DL codebases.

Good to have:

• Experience with MLOps, orchestration tools (e.g., Airflow, AWS Step Functions), containerization (Docker, Kubernetes.
• Exposure to optimization toolkits (ONNX, TensorRT) and serving frameworks (Triton, TorchServe).
• Experience with experiment tracking (e.g., Weights & Biases, Comet)
• Understanding of alignment techniques like RLHF or curriculum learning