Portrait of Jabavu Adams
Jabavu "Jaba" Adams

AI, Machine Learning, Neuroscience, Biophysics, Games, Teaching

Ollie: The Family AI Assistant

Confabulation Corporation · Employee #1, Staff AI Developer

Employee #1 (remote) at Confabulation Corporation — a US startup spun out of the Allen Institute's AI2 accelerator. The AI assistant demo I helped program (me + the CTO) was instrumental in closing a $5M seed round led by Khosla Ventures. From there my main focus was the Grocery List feature end-to-end, while jumping in to help on conversation engineering when the team needed extra hands or had a tricky issue to debug — typically getting external data into the LLM's context, or tracking down conversation failures that traced back to other parts of the system feeding the model incorrect or ambiguous information. Returned for a second stint to help the team through the transition as the original CTO departed.

Contributions

  • Seed-round demo. Co-programmed (with the CTO) the AI assistant demo that helped us raise the $5M seed round from Khosla Ventures. Backend stack: TypeScript, Node.js, Docker, Redis, Elasticsearch, Postgres.
  • Conversation engineering. Key contributor on context engineering — getting external data (Postgres, Elasticsearch) into the LLM's context for conversations, and systematically debugging conversation failures that traced back to other parts of the system feeding the model bad or ambiguous information.
  • Grocery List feature (lead). Built it end to end — turning a week of planned meals into a shoppable ingredient list. Python, DSPy.
  • In-house LLM evals. Built model testing and scoring tools (Excel, Python, Flask) so the team could compare prompts and model versions quantitatively.
  • Cross-team support. Dropped into other areas of the codebase as needed — extra hands wherever the team was stretched.
  • AI coding tools advocate. Championed AI-assisted coding tools and agents (Devin, Mentat, and others) — evaluating new tools, building team workflows around them, and helping teammates adopt them productively.
  • Mentorship. Mentored junior developers on software engineering and debugging distributed-systems issues.
  • Occasional DevOps. Infrastructure work across Linux, GitHub, Docker, GCP, Kubernetes, Temporal, and DigitalOcean.
  • Second stint — CTO transition. Brought back to help keep the team moving as the original CTO departed; ended up staying for over a year.
Visit try.ollie.ai →