Gen AI Engineer

• Design and implement stateful multi-agent workflows using LangGraph (checkpointers, retries, subgraphs, tool calling).
• Define Agent-to-Agent (A2A) interaction patterns for decomposition, verification, and self-correction.
• Build tool-using agents with structured outputs, schema enforcement, and deterministic execution paths.
• Handle agent failure modes such as hallucinations, tool misuse, and partial execution.
• Select and tune vector stores (FAISS, Milvus, Pinecone, Weaviate). Inference & Model Optimization
• Operate and optimize LLM inference pipelines with focus on latency, throughput, and cost.
• Work with vLLM (continuous batching, memory efficiency).
• Make informed trade-offs between model size, context length, and output quality.
• Apply quantization and other inference-time optimizations where required. Evaluation & Iteration
• Design and run LLM evaluation workflows using tools such as LangSmith, Ragas, TruLens, or equivalent.
• Define acceptance metrics for Grounded Ness, Context relevance, Answer quality.
• Use evaluation results to iterate on prompts, retrieval strategies, and agent design.
• Ability to reason about: Attention mechanisms and scaling, Decoder-only vs encoder decoder architectures o Prompting vs retrieval vs fine-tuning trade-offs.

Agentic Workflows (Strong), Communication & collaboration (Strong), Lang chain, Python (Strong), LLM Foundations. Good to have skills:

• DevOps - GenAI, Transformer-based models and seq-to-seq paradigms.
• Pharma industry/domain experience is preferred., * Hands-on experience solving non-trivial GenAI use cases. Agentic & RAG Expertise
• Proven experience building agentic workflows with LangGraph.
• Strong understanding of tool calling, structured outputs, and schema contracts.
• Deep experience with RAG systems, including retrieval evaluation and optimization.
• Experience with vector databases and embedding strategies. Inference & Evaluation
• Experience running and tuning LLM inference workloads.
• Familiarity with vLLM or similar inference engines.
• Experience with LLM evaluation frameworks and metric-driven iteration.