100Minds.ai
A practice-based leadership and power skills training platform powered by AI. I built the core AI infrastructure — a voice-enabled tutor, an interactive AI avatar, and RAG pipelines grounded in curated learning content.
The Challenge
Leadership training typically lives in static courses, PDFs, or expensive in-person workshops. The team at 100Minds wanted to make leadership skills genuinely interactive — where learners could practice real scenarios with an AI that responds intelligently, retains context, and grounds its answers in verified content rather than hallucinating advice.
The Solution
I built a RAG pipeline that ingests and chunks the platform's curated training content into a vector database, so every AI response is anchored to real material. On top of that, I integrated a voice-enabled tutor for spoken interaction and an AI avatar for immersive scenario practice — making the experience feel closer to a coaching session than a course.
What I Built
Designed and implemented the RAG pipeline — ingestion, chunking, embedding, and retrieval — grounding all AI responses in curated training content
Built the voice-enabled AI tutor with real-time streaming, low-latency response, and natural interruption handling
Integrated an interactive AI avatar for immersive scenario-based learning, synchronized with live audio output
Architected the LangChain/LangGraph workflow orchestrating context retrieval, response generation, and session memory
Optimized chunking strategy and vector retrieval to maximize factual precision without losing conversational context
The Story
The hardest part wasn't the RAG pipeline — it was making the voice interaction feel natural. Early versions had noticeable latency and awkward turn-taking. I had to tune the streaming pipeline and interruption handling to get it to a point where users forgot they were talking to an AI. The avatar added another layer of complexity: synchronizing lip movement and expression with live audio output required careful orchestration between the voice model and the avatar rendering layer.
What I Learned
RAG quality is entirely dependent on chunking strategy — chunking too coarsely loses precision, too finely loses context. I spent more time on this than any other part of the system. I also learned that voice AI UX is its own discipline: latency tolerance, interruption handling, and conversational pacing matter as much as the model quality underneath.
Technologies Used
Back to
All Case Studies