NMNick McCarthy

Experience

Professional and education timeline.

Each entry expands with skills, related artifacts, and the context behind the role.

ProfessionalEducation

Skills & technologies used

Amazon BedrockAmazon SageMaker AIAmazon Bedrock Custom Model ImportAmazon Bedrock Reinforcement Fine-TuningModel customization workflowsLLMOpsReinforcement learning with verifiable rewards (RLVR)OpenAI-compatible APIs
Amazon Bedrock RFT workflowBedrock Custom Model Import performance chartsVLM document-to-JSON fine-tuning workflowMulti-provider GenAI gateway architecture
  • Worldwide technical lead for OpenAI on Amazon Bedrock, driving a multi-billion dollar annual revenue target — aligning OpenAI, the Bedrock service team, and the worldwide SA community as the highest point of technical escalation to unblock customers and accelerate adoption.
  • Leading the OpenAI Technical Field Community and Champions Group to enable 1,000 field SAs — designing and delivering three two-day train-the-trainer bootcamps in New York, Seattle, and London covering OpenAI models on Bedrock and OpenAI Codex enterprise deployment; graduates now run 12 local, geo-specific bootcamps, scaling enablement across NAMER and EMEA ahead of GA.
  • Drove 29 influenced opportunities worth $16M in ARR, 11,326 registrations, and ~30 lighthouse enterprise customers through the AWS AI League go-to-market — by building hands-on model-customization challenges, customer GTM assets, launch workshops at the AWS NYC Summit and re:Invent, a 16-company hackathon across five time zones, and the Atos customer story.
  • Shaped the launch readiness of SageMaker AI's serverless, agentic model-customization capability as a key scientific advisor — contributing RLVR workflow design, HyperPod training recipes, a penetration-testing playbook, Robin AI preview support, and a 30-person cross-org preview program — by partnering with the Amazon Bedrock Applied Science team to turn new capabilities into field-ready guidance.
  • Published a reusable LLMOps fine-tuning platform (the Model Customization Platform Accelerator) to AWS Samples and delivered it as a re:Invent Builder's Session and AWS North America Tech Summit session — giving customers and field teams a production-ready blueprint for fine-tuning on SageMaker AI.
  • Drove Talent.com's foundation-model adoption decision and co-authored AWS's reference guidance on a mature GenAI foundation — including the Generative AI Lens multi-tenant scenario and model-customization readiness frameworks — by running tailored SageMaker AI and Bedrock workshops adopted across customer engagements.
  • Built the internal OpenAI and Anthropic model parity dashboards used by the Bedrock field — running load testing and performance benchmarking across the model families and Codex so SAs can position Bedrock deployments against first-party provider offerings with real data.
  • Driving the AI-DLC (AI-Driven Development Life Cycle) with Codex workstream — applying AWS's AI-DLC methodology to OpenAI Codex on Bedrock for enterprise agentic development, with a joint blog post publishing soon.
  • Presented joint AWS and OpenAI breakout sessions at the AWS New York and Los Angeles Summits on OpenAI models and Codex on Amazon Bedrock, a 250-person AWS London Summit breakout, and a re:Inforce chalk talk, and led a six-person re:Invent Builder's Session team for an LLMOps platform build-out.
  • Turned launch collaborations into public, field-ready guidance — publishing reinforcement fine-tuning on Amazon Bedrock walkthroughs informed by RLVR preview work with customers like Robin AI, VLM fine-tuning for document-to-JSON workflows, Bedrock Custom Model Import, and multi-provider GenAI gateways.

Skills & technologies used

LLMOpsAmazon SageMaker AIAmazon BedrockAWS TrainiumAWS InferentiaResponsible AIModel evaluationFine-tuning workflows
Time series forecasting with Amazon SageMaker AutoMLBMW cloud efficiency analytics dashboard
  • Led delivery of an end-to-end LLMOps platform for Booking.com — building automated fine-tuning, evaluation, load-testing, and deployment workflows on Amazon SageMaker that gave their ML teams a repeatable path from experiment to production for open-source LLMs.
  • Took Booking.com's AI Trip Planner to production, serving 100M+ app users — by fine-tuning open-source LLMs for destination recommendations with the AWS Generative AI Innovation Center, reducing the product's dependency on proprietary OpenAI models and giving Booking.com control over cost, latency, and model behavior.
  • Led the winning team at the Aviva GenAI Hackathon 2023 as sole technical advisor and ML engineer — designing and building AutoClaim, an automated insurance-claim bot powered by Amazon Bedrock and Amazon Kendra.
  • Defined go-to-market strategy and scoped GenAI engagements for major energy enterprises including TotalEnergies, ENI, and BPX Energy — serving as GenAI technical lead for AWS Professional Services Energy UKIR and delivering 1:1 C-level advisory through the UKIR GenAI Solutions Launchpad.
  • Shaped the roadmaps of four AWS GenAI services — Amazon SageMaker AI, Amazon Bedrock, AWS Trainium, and AWS Inferentia — by channeling enterprise customer feedback into feature requests, Responsible AI governance, and beta testing of preview features such as Bedrock Custom Model Import.
  • Published public work on time series forecasting with Amazon SageMaker AutoML and BMW cloud efficiency analytics with Amazon QuickSight and Amazon Athena.

Skills & technologies used

MLOpsAmazon SageMaker AIAmazon SageMaker ClarifyAmazon SageMaker CanvasExplainable AICI/CD pipelinesGenomics workflowsData migrationsPartner enablementPre-sales advisory
SageMaker Canvas facies classification workflowBundesliga xGoals SHAP pressure plots
  • Influenced $31M in AWS Professional Services bookings across 23 unique opportunities — through strategic pre-sales discussions with C-level stakeholders.
  • Cut model deployment time 30% and improved model accuracy 60% for BPX Energy, unlocking ~40% ROI-increase potential — by implementing an MLOps framework on Amazon SageMaker AI with automated monitoring and re-training; the platform was showcased by the BPX CTO at re:Invent 2022.
  • Cut ML use-case onboarding time 83% and saved $453K in deployment costs for Ericsson — by leading delivery of an MLOps platform on Amazon SageMaker AI that took onboarding from three days to four hours across four production ML use cases.
  • Migrated 300 production apps to Aurora Serverless v2 ahead of schedule as Lead Data Architect for Siemens Mendix — delivering 3x faster large scaling events and ~30% database cost savings, plus a product feature request adopted onto the Amazon RDS roadmap.
  • Saved €200K across 10 customer engagements for AWS partner ML6 — by designing a code-promotion CI/CD pattern that moves ML code between environments rather than promoting opaque model artifacts.
  • Trained 175+ consultants across five AWS partners — by co-founding the Partner Workshop Factory, a 19-person team that built 14 reusable workshops, leading EMEA partner ML immersion days, and organizing two GenAI hackathons for Rackspace.
  • Mentored 89 colleagues to AWS speaker certifications as a UKIR Certification Champion, coached two graduates into the ML Technical Field Community, and co-created the Artifact Acceleration Day — guiding 96 graduates to 25 published AWS guidance artifacts.
  • Cut deployment time 87% and saved AstraZeneca $490K annually plus 120 scientist-hours per year — by designing a new deployment workflow, a Genomics Data Lake, and the AI Bench MLOps platform with five SageMaker Pipeline products, enabling 300+ researchers to manage 1,200+ ML experiments annually.
  • Delivered the explainability layer for Deutsche Fussball Liga's in-production expected-goals models — designing Amazon SageMaker Clarify experiments published as an AWS ML Blog post and a German DevOps magazine feature, and surfacing a missing SHAP KernelExplainer capability as a product feature request.
  • Built AWS Dash, an internal KPI tracker used by 3,000+ consultants and managers — leading a four-person backend team on API Gateway, Lambda, Glue, and RDS.
  • Presented externally at the AWS Manchester User Group and UCISA23, and created the AI Air Hockey session at the re:Invent 2022 Builder's Fair, attended by 250 participants.

Skills & technologies used

Proximal Policy Optimization (PPO)Deep Deterministic Policy Gradient (DDPG)Soft Actor-Critic (SAC)Portfolio constructionMacroeconomic dataReinforcement learning trading environments
  • Integrated PPO, DDPG, and SAC into a trading platform that mapped investment signals to target portfolio weights.
  • Designed RL trading environments and researched neural architectures, action distributions, and macro-data state representations.
  • The resulting models delivered superior out-of-sample returns versus xAI's legacy long-only strategies and moved into production on the platform.

Skills & technologies used

Deep reinforcement learningNatural language processingPortfolio optimizationPythonAcademic research writing
Machine Learning MSc visual
Chess pieces from MIT Technology Review SentiMATE article
  • Thesis: Deep Reinforcement Learning for Portfolio Construction & Optimisation, in collaboration with xAI Asset Management.
  • Completed the "Deep Learning and Reinforcement Learning" module, taught by Google DeepMind at UCL.
  • Related public research: SentiMATE, accepted for oral presentation at AIIDE 2019 and covered by MIT Technology Review.

Skills & technologies used

Experimental physicsDoppler-free spectroscopyOpticsUncertainty analysisData analysis
UCL Wilkins Building
  • Thesis: Investigation of the hyperfine structures of Rubidium using Doppler-free, frequency-modulated saturated absorption spectroscopy.
  • Graduated with First Class Honours.