Sr. Applied Scientist - Agents

AuxoAI is hiring a Senior Applied Scientist to design and deploy production-grade AI agents capable of structured reasoning, planning, and decision-making.
This role focuses on building intelligent agent systems that combine LLM-based reasoning with classical planning, search algorithms, and optimization techniques. The ideal candidate will develop robust agent architectures that operate reliably in real-world environments with constraints around latency, cost, uncertainty, and limited context windows.
You will work on advanced AI systems that power autonomous workflows, decision engines, and tool-driven agent ecosystems. You will work on problems where existing architectures may not be sufficient, and you will be expected to experiment with new approaches that combine machine learning, graph algorithms, and classical AI techniques to build reliable, production-grade systems.
Responsibilities:
Design and architect modular AI agent frameworks incorporating skill decomposition, tool orchestration, and persistent state tracking.
Implement planning and search algorithms such as Monte Carlo Tree Search (MCTS), beam search, A search, heuristic search, and graph-based planning approaches to support complex decision-making tasks.
Develop decision-making loops that balance trade-offs between exploration vs. exploitation, cost vs. accuracy, and latency vs. reasoning depth.
Build structured memory systems including episodic memory stores, semantic memory layers, and vector-based memory with optimised retrieval strategies.
Design tool-calling architectures with strong execution validation, retry mechanisms, and failure recovery strategies.
Develop evaluation frameworks to measure agent performance using task success metrics, rollout simulations, and multi-sample validation approaches.
Improve agent performance through techniques such as distillation, synthetic trajectory generation, prompt compression, and context pruning.
Deliver production-ready agent systems that meet operational requirements around reliability, cost efficiency, throughput, and observability.
Requirements:
2-5 years of experience building machine learning or AI systems in production environments.
Strong experience implementing search or planning algorithms beyond basic use cases, including tree search or heuristic-based planning approaches.
Hands-on experience with Monte Carlo Tree Search (MCTS) or related decision-making frameworks.
Strong understanding of state-space representations, heuristic design, and decision boundary trade-offs.
Experience building or extensively customising agent frameworks for real-world applications.
Hands-on experience designing tool-use or function-calling architectures under practical system constraints.
Strong Python engineering skills with a focus on scalable and reliable system design.
Nice to Have:
Experience with reinforcement learning techniques such as policy gradients, value estimation, or reward modelling.
Experience building multi-agent or collaborative agent systems.
Experience designing evaluation frameworks for agent robustness and reliability.
Experience optimising LLM inference pipelines for latency, throughput, and cost efficiency.
Familiarity with distributed task orchestration systems and large-scale AI workflow management.

Experience
3-6 yrs