HE

Harry Ephremsen

+44 7908 467 167 | harryephremsen@me.com

Professional Summary

AI-native builder and technical systems architect with expertise in end-to-end product development. Experienced in investment analysis, systems architecture, and technical delivery.

Employment

Technical Systems Architect - ID Theory

May 2024 to Present
  • Architected and built a comprehensive internal data infrastructure supporting investment, risk, and operational workflows.
  • Developed full-stack portfolio analytics dashboards, APIs, and research AI agents for fund automation.
  • Built large-scale asset-scanning systems identifying emerging sectors, narratives, and quantitative signals.

Investment Analyst - ID Theory

August 2022 to May 2024
  • Sourced and conducted due diligence on venture/liquid investments with focus on AI, autonomous agent infrastructure, and machine learning.
  • Led technical analysis on AI/ML-driven investment opportunities focused on distributed compute, inference, training, and agent coordination.
  • Lead on mutual fund committee seats (e.g. Seedclub AI Ventures & Metropolis AI DAO).

Operations Executive - ID Theory

August 2021 to February 2023
  • Designed and maintained portfolio management systems in Excel (VBA) and Power BI for a multi-chain portfolio.
  • Led two annual audits end-to-end as primary liaison to third-party administrators and auditors.
  • Managed monthly NAV reconciliations, investor reporting, and operational workflows.

Projects

myaicards(Independent Project)
  • AI-powered greeting card platform (in beta testing) using image generation models to transform personal memories into printed cards.
  • Built end-to-end: concept to production including microservices architecture handling image generation, ordering, payment, and print-on-demand fulfilment.

LLM Evaluation Platform

(Independent Project)
  • Blind, automated LLM benchmark comparing model capabilities across major providers.
  • Implemented anonymised outputs, peer LLM cross-scoring, and aggregated rankings with analytics dashboard.
  • Operationalised research paper on human vulnerability to AI into a web platform with a dynamic questionnaire.
  • Implemented 0-100 AI Vulnerability Score with analytics for user benchmarking and large-scale data collection.

Technical Skills

AI Tools:

Claude Code, Codex, MCP servers

Languages:

Python (backend, analytics, ML), TypeScript/JavaScript (frontend, Node.js)

Infrastructure:

AWS, Google Cloud, DigitalOcean, Vercel, Docker, CI/CD, Git, Ansible, Kubernetes

Databases:

PostgreSQL, MongoDB, Redis

Web & APIs:

REST APIs, WebSockets, Webhooks, RPCs

OS:

Arch Linux (daily driver), Ubuntu (servers), macOS (mobile)

Education

Machine Learning & Data Science - Le Wagon

May 2024 to November 2024
  • Intensive program covering supervised/unsupervised ML, deep learning, and model deployment.
  • Developed ResNet-based binary classification model for mushroom toxicity detection on Kaggle dataset (20.5k training, 5.2k validation, 6.4k test images) achieving 90.57% accuracy and 0.94 AUC.

Additional

Writing: Co-author of "The Post-Human Economy"
Interests: Endurance sports (Ironman 70.3 Mallorca competitor), golf, skiing, PC tinkering, linux ricing, AI developments, technology.