Advisory Research Software Engineer

Advisory Research Software Engineer, IBM Corporation, Cambridge, MA (Up to 40% telecommuting permitted):

• Design and implement AI systems that address complex SQL and data transformation tasks requiring multi-step reasoning beyond single-pass generation approaches.
• Lead the technical development of the Query Planning Gym (QueryGym) framework to support structured reasoning and evaluation on enterprise-scale workloads.
• Architect agent-based workflows within QueryGym to support planning, validation, critique, and repair as coordinated system components.
• Integrate execution-aware tools into agent workflows to improve correctness, reduce retries, and optimize system efficiency.
• Translate research concepts in reasoning and planning into product-compatible architectures, develop production-grade code, testing systems, benchmarking performance, and analyzing results.
• Integrate and analyze benchmarks such as ELTBench against QueryGym for evaluating complex, real world SQL transformation workloads.
• Develop algorithms and agentic systems that improve LLM reliability.
• Collaborate with product teams to analyze production data transformation pipelines and identify schema drift and execution issues.
• Develop remediation approaches that translate plan level failures into actionable configuration or transformation guidance for users, including validation using benchmarks.
• Work on design and implementation of autonomous agents for AppWorld.
• Prepare technical documentation, design specifications, and implementation guidance for downstream teams to operationalize developed systems.
• Prepare research publications and end-to-end demonstrations to communicate system capabilities to internal and external audiences.
• Design and implement a synthetic dataset and benchmark generation pipeline for long, complex SQL queries.
• Utilize: Python, Transformer Architecture, Data & AI LLMs (Fine Tuning, Inference, Evaluation), Container Orchestration Platforms (Kubernetes, Docker, OpenShift), Data & AI Microservice deployment, Agentic AI frameworks, Natural Language Processing (NLP).

Required: Masters degree or equivalent in Computer Science or related (employer will accept a Bachelor's degree plus five (5) years of progressive experience in lieu of a Masters degree) and one (1) year of experience as a Research Software Engineer or related. One (1) year of experience must include utilizing Python, Transformer Architecture, Data & AI LLMs (Fine Tuning, Inference, Evaluation), Container Orchestration Platforms (Kubernetes, Docker, OpenShift), Data & AI Microservice deployment, Agentic AI frameworks, Natural Language Processing (NLP). $203010 to $239300 per year. Full time. AV217.