AI Safety · Platform Engineering · Social Impact
Human-Centred AI:
Safety is Not a Feature.
Engineer and founder at the intersection of AI safety, platform engineering, and social impact. Primary area of interest in agentic development and collaboration to create tangible impact within the sector of AI safety.
Learn more →Research & Areas of Interest
Safety & Alignment Lab
Unified research addressing ethical implications, safety concerns, policy frameworks, and community impact of AI systems. A holistic approach to responsible AI through integrated tracks.
Energy Transition Lab
Investigating how AI can accelerate the shift to sustainable energy systems. Research into grid optimisation, demand forecasting, and climate-aware infrastructure planning.
Creative AI Lab
Exploring the intersection of generative systems and human creativity — from co-authorship to novel art forms. Building tools that augment, not replace, creative expression.
Developer Tools & Frameworks
Creating practical, open-source tools and frameworks for responsible AI development. Democratising access to ethical AI tooling through transparent, collaborative builds.
Tools
Open-source tools and demos being built at HK Murphy — at the intersection of AI safety engineering and ML security.
LLM Safety Eval Harness
A harness engineering approach to LLM safety evaluation — wrapping models in reproducible, composable test scenarios rather than one-off scripts. Covers prompt injection resistance, refusal consistency, output sanitisation, and instruction-following under adversarial conditions. Harnesses are first-class artifacts: versioned, shareable, and independent of the model under test. Built on UK AISI's Inspect framework.
ML Security Demo Harness
Harness-based demonstrations of ML security attack surfaces: adversarial examples, data poisoning, model extraction, and membership inference. Each attack scenario is encapsulated as a standalone harness — reproducible, self-documenting, and runnable against any compatible model. Designed as both a security education tool and a template for building your own security evals.
Agentic Safety Monitor
Lightweight observability layer for agentic AI systems — tracking tool call sequences, detecting anomalous behaviour patterns, and flagging potential safety violations in autonomous pipelines. Informed by production agentic AI deployment at The Economist and the emerging literature on agentic failure modes.
RAG Security Harness
Security testing harness for RAG pipelines — systematically probing retrieval poisoning, indirect prompt injection via documents, and output exfiltration vectors. Harnesses are scoped per attack class, composable into full pipeline audits, and designed to run in CI alongside functional tests. Extends the AI Engineering course RAG work with a dedicated adversarial layer.
About
Engineer, founder/partner, and researcher deliberately transitioning toward AI safety — the most pressing problem of our time.
I'm a platform engineer and technical leader currently serving as Co-Technical Lead for an Agentic AI project at The Economist, where I've gained firsthand exposure to the safety challenges of deploying autonomous AI systems in production. That work has deepened both my understanding of the problem and my conviction that my background can contribute meaningfully to solving it.
Before AI, I founded and grew social enterprises in Guatemala that reached over 4,000 people — work recognised with an Ashoka Fellow nomination, a SOCAP Programme nomination, and three accelerator wins across IMPULSA National, Alterna, and Booking Booster. That work instilled a discipline of evaluating decisions by expected impact rather than convention. I'm now applying that same framework to the question of where a platform engineer with 10+ years of technical leadership in Platform, Agentic coLead, DevSecOps and Data, statistics training, and a linguistics background can have the greatest effect on AI safety.
I'm actively upskilling in ML and exploring whether my comparative advantage lies in AI safety engineering — contributing immediately to infrastructure and security — or in the longer investment of AI safety research via DPhil or intensive fellowship. hkmurphy.com is the platform through which I'm building, researching, and connecting in public.
Focus Areas
Experience & Education
Experience
Technical Co-Lead, Agentic AI — IndexAI
The Economist
Co-leading technical delivery of IndexAI — The Economist's agentic AI product built on AWS Bedrock AgentCore. Directly responsible for the safety, reliability, and engineering architecture of an autonomous AI system in production. Firsthand exposure to the alignment and safety challenges that arise when deploying agentic systems at scale.
Engineering Lead, Platform Engineering
The Economist Group
Engineering leadership across The Economist and EIU platforms, including agentic AI development. Gained firsthand exposure to the safety challenges and failure modes of deploying autonomous AI systems in production — deepening both understanding of the problem and conviction that this background can contribute meaningfully to solving it.
Engineering Lead, DevSecOps & SRE
The Economist / EIU
Led DevSecOps enablement and site reliability engineering across The Economist Group. Built security-first infrastructure underpinning global media operations, developing practices and a security-first perspective that now informs AI safety thinking.
Site Reliability Engineer, Full Stack Engineer — Big Data Project
Economist Intelligence Unit
SRE and software engineering on the Viewpoint Big Data project. Foundation in production reliability, observability, and large-scale data infrastructure. Also completed School of Code bootcamp (2020).
Head of Social Enterprise & Board Member
Niños de Guatemala
Grew and ran multiple social enterprises as a sustainable income source facilitating education for 525 children and ~4,000 community members. Ashoka Fellow Nominee 2019, SOCAP Nominee, ALTERNA Seed Capital Winner, IMPULSA National Winner 2018, Tour Operator of the Year 2017, 2018 & 2019 (Luxury Travel Guide), Antigua10x Business Incubator Finalist. Pitch presenter at FLII 2019 (LatAm Forum for Impact Investment, ~1,000 attendees).
Strategy & Growth Lead / Digital Product Specialist
LIFULL Connect (formerly Trovit)
Post-acquisition growth leadership across APAC & EMEA for a platform spanning 250 sites, 63 countries, 300M ads/month and 180M visits/month. Led product rollout of Real Time Bidding technology; drove ~€36M p.a. revenue growth. First point of contact for 16 Country Managers.
Emerging Markets Lead / Content & Data Lead
Trovit
New market launches and business development across emerging markets at a Barcelona-based tech company. Organised Trovit Talks — monthly events for 200+ tech professionals with speakers from SeedRocket, 4 Founders Capital, and others.
Education
MSc Data, Economics & Development Policy · MSc Statistics & Data Science
Massachusetts Institute of Technology
Dual postgraduate programme covering data science, statistical methods, and development economics.
Entrepreneurship in Emerging Economies
Harvard University
Executive education programme focused on entrepreneurship and venture creation in emerging market contexts.
BA French & Italian Studies — Translation & Interpretation · Politics & Education
University of Warwick
Major in Translation & Interpretation; Minor in Politics & Education. Winner of URSS Research Bursary: Mobility & Identity Formation.
Events
Applying to Attend
EA Global: London 2026
InterContinental London — The O2 · London, UK
Attended
NVIDIA GTC 2026
San Jose McEnery Convention Center · San Jose, CA
Attended
AI Growth Summit
London · London, UK
Attended
Gen AI Summit
London · London, UK
Attended
AI Advantage Summit
London · London, UK
Let's build
together.
Open to AI safety engineering roles, research collaborations, fellowship conversations, and anyone thinking seriously about where technical talent can have the most impact on AI risk.
GitHub
impactyogi ↗Status
Open to AI Safety collaborationLocation
Oxford, UKSacatepéquez, Guatemala