Systems

Small honest demos, explicit evidence, clear scope.

These projects are framed by the problem they address, the capability they demonstrate, the evidence surface behind them, and the next validation layer. Clinical and regulated systems stay synthetic, staff-review oriented, and deliberately bounded unless a separate governance path exists.

Clinical workflow support Synthetic workflow prototype

UroPrevisit Navigator

A urology previsit workflow that uses synthetic cases, adaptive governed questions, missing-field repair, and clinician-review summaries for PSA follow-up and CRM-ready proposal work.

Versioned synthetic-case workflow with patient, nurse, clinician, visit-packet, and reviewer views.ASR-ready adaptive-question path that stops before diagnostic drift.
Market demo / staff-review support Public-safe demo

AI Triage Kiosk Demo

A synthetic vital-aware kiosk intake demo that returns governed English follow-up questions and a staff-review summary for an imedtac market demonstration.

Two-endpoint API and field semantics aligned for the June demo baseline.Runtime checks for demo-only safety boundaries and governed-question ranking.
Voice AI Realtime prototype

Jarvis Voice Sight

A mock-first realtime voice agent prototype with always-listening interaction, VAD, barge-in, turn isolation, sentence-level streaming TTS, and Ollama/vLLM runtime support.

Realtime mode, stale-audio discard, cancellable TTS queue, and sentence-level audio streaming.Mock mode for microphone-unavailable contexts while preserving the full pipeline contract.
Local meeting-summary trust Local desktop workflow

Project AURA

A local-first meeting-summary workflow that turns corrected transcripts into structured reports while making runtime model readiness and error states visible.

Parallel layered extraction, field schemas, prompts, sample report, and tests.Local Ollama preflight with exact model-tag checking and user-confirmed model pull behavior.
AI systems Prototype

Speech Evidence Intelligence Pipeline

An evidence-aware speech intelligence pipeline using ASR, retrieval, and LLM extraction to turn long-form conversational audio into structured, reviewable outputs.

ASR and transcript segmentationRetrieval over evidence-bearing spans
Fraud analysis Research seed

Fraud Conversation Analysis with RAG

A research-led case study on retrieval-augmented fraud conversation analysis designed to keep LLM outputs grounded in transcript evidence.

Transcript indexingEvidence-bearing dialogue retrieval
Cybersecurity Research seed

Federated Learning Leakage Study

A research case study on federated learning privacy leakage, gradient inversion risk, and defense trade-offs for sensitive collaborative training.

Threat-model framingGradient leakage literature