You are viewing a preview of this job. Log in or register to view more details about this job.

AI Algorithm Intern

Mi Terro
Email

William (a subsidiary of Mi Terro Care) is an AI companion for emotional wellbeing. Not a chatbot. Not a meditation app. An AI that actually understands you, by learning from your conversations, your screen time, your sleep, and even what you tell ChatGPT and Claude when you think nobody’s listening.

We’re building the first product that stitches together signals from every corner of someone’s digital life and synthesizes them into one emotional profile.  The result: an AI that notices your anxiety is rising before you do.

We’re a lean, fast-moving team. We ship daily.  We don’t do meetings about meetings. 

Why Join Us

  • You’ll build production features that real users interact with from day one, not internal tools nobody sees.
  • You’ll work directly with the founder/CEO. No layers. Your ideas go from conversation to production in hours.
  • You’ll gain hands-on experience with cutting-edge AI agent frameworks (OpenClaw, Harness Agent, LangChain, CrewAI) before most engineers even know they exist.
  • You’ll learn how to build and launch a real world product, from AI infrastructure to growth hacking to customer acquisition.

We move fast and trust our people. If you prove yourself, there’s a clear path to a full-time role with equity.

 

AI Algorithm Intern

What You’ll Build

The behavioral intelligence engine that makes William smarter than a generic chatbot. William’s 5-layer architecture processes raw signals (app usage, sleep, screen time) through local classification, stores them in a markdown vault, and generates personalized coaching via multimodels. You’ll improve every layer.

 

Specific Projects

  • Fine-tune the emotional signal extraction pipeline, improve how William detects anxiety, burnout, and relationship stress from user chat history and AI sync imports
  • Build the composite health indicator engine, combine weight trends, restlessness, sleep duration, breathing rate, and asymmetric pressure into a single wellbeing score
  • Implement memory vault compaction, when the user has 500+ memories, intelligently merge and summarize without losing emotional nuance
  • Design the proactive outreach algorithm, William should reach out before users spiral, not after. Build the trigger system that decides when and what to say
  • Optimize the system prompt builder (behaviorEngine.js). the context window is limited; decide what memories, patterns, and AI sync insights are most relevant for each conversation
  • Build cross-platform behavioral correlation, connect desktop app usage patterns with mobile mood entries and AI sync emotional signals

 

You’re a Fit If

  • You’ve fine-tuned or prompt-engineered LLMs for real applications (not just coursework)
  • You understand embedding spaces, RAG pipelines, and context window management
  • You can write production Python or JavaScript, not just notebooks
  • You’ve worked with OpenAI API, Anthropic API, or open-source models (Gemma, Llama, Mistral)
  • You’ve read papers on emotion detection, sentiment analysis, or behavioral prediction

Bonus: experience with on-device inference (Ollama, llama.cpp, ONNX Runtime)
 

How to Apply

We don’t care about your GPA, your school, or how many buzzwords are on your resume. We care about what you’ve built and how fast you can execute.

Join our Discord channel. Then send us:

  • A short intro (who you are, which role you want, why this excites you).
  • Links to things you’ve built (GitHub, live projects, portfolio, growth metrics, anything).
  • Your available hours per week and earliest start date.

Email: hello@iamwilliam.xyz

We review applications on a rolling basis. If you’re good, we’ll reach out to you very soon.