• Own the AI execution and infrastructure layer used by research and product teams • Design and build high-performance Python systems for: - scalable model inference - training orchestration - large-scale data processing • Partner closely with research and backend engineers to productionize models and expose them via APIs • Design and operate distributed pipelines and task queues for batch and streaming workloads • Optimize GPU inference for latency, throughput, and cost efficiency • Own the MLOps lifecycle, model deployment and versioning; monitoring and alerting • Build and maintain CI/CD pipelines for services and ML workflows • Debug and resolve performance bottlenecks across Python, GPUs, networking, and storage • Contribute to infrastructure design decisions and long-term architecture

• Design system architectures involving LLMs, intelligent agents, and multi-agent systems; • Develop end-to-end Generative AI pipelines, from experimentation to production deployment; • Build autonomous agents, custom tools, functions, and integrations with external APIs; • Implement orchestration between agents, memory systems, expanded context windows, and decision-making mechanisms; • Conduct evaluation, monitoring, and observability of models (RAG, agents, prompts, and behaviors); • Optimize cost, latency, and quality of production solutions; • Work with prompt versioning, automated testing, and continuous validation; • Collaborate with product, marketing, journalism, analytics, and data engineering teams to evolve the company’s Generative AI platform.

• The focus is on implementing autonomous reasoning flows, integration via MCP, and control of non-deterministic state. • **Orchestration of Agentive Flows:** Build loop logics (Think, Act, Observe) for complex Supply Chain tasks (stockout analysis, recalculating Sell-In) using frameworks such as **LangGraph, CrewAI, Agno or Bedrock AgentCore**. • **Implementation of MCP Servers:** Develop and expose tools via *Model Context Protocol*, creating secure bridges (APIs) between LLMs and **PostgreSQL, Iceberg** databases and Rule Engines. • **Reasoning and State Engineering:** Manage short- and long-term memory and implement **Thought Signatures** to ensure cohesion in multi-step executions (Gemini 3 and Claude 3.5). • **Tool/Function Calling Optimization:** Design strict contracts using **JSON Schemas**, implementing fallback and retry logic to mitigate hallucinations in external calls. • **AgentOps Discipline:** Ensure full AI observability (tracing, latency, token consumption, and decision paths) via **LangSmith, Phoenix or AWS CloudWatch**. • **Guardrails and Security:** Configure automated barriers (e.g., **Amazon Bedrock Guardrails**) to intercept hallucinations and block unauthorized actions on master data. **

• Conduct experiments with LLMs and explore different architectures and techniques such as SFT, RLHF, Adapters, LoRA, and related approaches to improve model capabilities • Develop and maintain evaluation methodologies and frameworks to measure model quality, accuracy, and user satisfaction using both offline and online metrics • Optimize models for inference efficiency, speed, and scalability to meet production requirements • Collaborate closely with data scientists, software engineers, and product managers to integrate AI solutions into the product pipeline