DevSight

The problem

Investigating a new deviation meant manually keyword-searching four years of Trackwise records — roughly 30 minutes per question, and brittle: the right precedent was easy to miss if it used different wording. Investigators needed semantic recall over 2,213+ resolved cases without leaving the 30-day closure window.

Architecture

What it does

DevSight (formerly TrackLA) is a hybrid-RAG chatbot over four years of Trackwise deviations (2022–2026): 2,213+ resolved cases chunked into 31,577 searchable units. It replaces ~30-minute manual keyword searches with ~2-minute semantic ones — a 93% time reduction — and surfaces the actual precedent cases behind every answer.

Two-stage hybrid RAG

Search runs in two stages: a date-metadata filter narrows the candidate set, then semantic vector search ranks across 10 Trackwise fields. Date filtering happens post-search client-side — a deliberate workaround for Databricks Vector Search not handling DATE range operators cleanly.

• Conversation memory across the last 10 turns
• Real-time token streaming with voice input and read-aloud
• Feedback capture after a 50-second countdown
• Document upload limited to Word + Excel (image/PDF intentionally disabled to preserve streaming)

Result

Mobile-optimized with a fixed input bar, DevSight is one of three flagship enterprise apps (with AMIRA and BioTrack) that share the R-first analytics discipline — R does the deterministic work, the LLM only narrates.