Senior AI Engineer (Innovation and R&D)
Posted Updated
Description
- Researches, prototypes and proves new GenAI and agentic ideas that raise Equinix’s internal productivity and create capabilities for external products
- Works in the AI Innovation and R&D pod, running experiments from hypothesis to working prototype, measuring each against quality, cost and business value, and handing validated ideas to product and platform teams as production-ready designs
- Innovation areas include small language models (SLMs) and open-weight model customization, token economics, agent and workflow automation, and other emerging AI techniques
- Applies deep working knowledge of modern AI techniques: LLM and SLM development, parameter-efficient tuning, distillation, quantization, retrieval, agent orchestration, and tool-use and function-calling models
- Understands how model size, context length, caching and routing drive token cost and latency
- Reads and applies current research, and benchmarks open-weight and commercial models on the same tasks
- Applies established engineering and model-governance practices and proposes improvements
- Works on open-ended problems where the right approach is not yet known
- Frames each idea as a testable hypothesis with success metrics, selects methods, datasets and evaluation criteria, and decides with evidence whether to scale, change or stop
- Resolves tradeoffs among quality, cost, latency, security and operational complexity
- Work is reviewed at design and milestone gates
- Partners with Product, Platform, InfoSec, Legal, Finance and business teams to find high-value use cases
- Explains results and tradeoffs to senior and non-technical stakeholders
- Advises teams on which AI approach fits their problem
- Accountable through the pod for delivery of innovation targets
- Identify and test GenAI and agentic ideas for internal productivity and external product use cases
- Build prototypes quickly, validate them with business teams against defined success metrics, and hand off production-ready designs and reference implementations
- Track new models, techniques and tools; document which apply to Equinix and which do not
- Benchmark open-weight and commercial models on shared task datasets: quality, latency, cost, license terms
- Fine-tune, distill and quantize models for defined tasks, including tool-use and agentic behavior, and show measured gains over the baseline
- Feed results into the recommended-model catalogue owned by the Principal PM
- Build cost models per agent and workflow: tokens, calls, caching, compute and cost per completed task
- Test cost controls: model routing, context compression, caching, batching and budget limits
- Optimize serving (e.g., vLLM, SGLang) with the platform team
- Make experiments reproducible: versioned data, code, configs and results
- Gate each prototype’s promotion on evaluation thresholds, using the Agent Intelligence evaluation platform
- Turn failures and feedback into tuning data, regression tests or design changes
- Document each model or solution: intended use, data provenance, limitations and evaluation results
- Apply license, data-boundary and security requirements with InfoSec and Legal
- What success looks like:
- Each quarter, a set of innovation ideas is tested, with a documented scale, change or stop decision for each
- Each tested workflow has a measured cost per task and a documented model choice
- Every promoted prototype has an evaluation record and documentation
- Ability to build cost and performance models for AI workloads
- Track record of taking AI ideas from prototype to production or to a documented decision to stop
- Proficiency in Python and PyTorch, and with Hugging Face tooling
- Hands-on experience with at least one of: fine-tuning or parameter-efficient tuning (e.g., LoRA), distillation, quantization, agent or retrieval systems
- Experience building evaluation harnesses and using results to make decisions
- Clear written and verbal communication with non-technical stakeholders
- 5–8 years in ML or software engineering, including 3+ years in LLM or deep-learning development
- BS or MS in computer science, machine learning or a quantitative field, or equivalent experience
- Published research, patents or open-source contributions
- Experience with open-weight model licensing, model governance or responsible-AI programs
- Experience with multi-agent systems, model routing and MLOps
- Preference optimization or reinforcement learning, and tool-use or function-calling fine-tuning
- Experience with inference and serving (e.g., vLLM, SGLang), batching and GPU optimization
Skills
Sign in to read and join the discussion.