A curated selection of shipped tools, applied AI systems, and demos. Tip: press / to search.
Fast, infinite-scrolling video grid with cached manifest, thumbnails, overlay player, and auto-refresh when new clips land.
USD-quoted crypto board with paper trading, TP/SL, and trailing take-profit logic. Donchian + RSI helpers and a tiny HTTP API for your admin.
Complete shop platform: customers, vehicles, work orders, invoicing, tire quotes, accounting (P&L, expenses), and SIP phone integration with auto-popup customer lookup.
Web-based .rdp builder with presets, smart sizing, gateway, and one-click download/copy for fast remote access.
Technician workforce analytics for multi-bay shops: utilization, overtime forecasting, certifications, recruiting, and compliance—built to cut labor waste and keep bays turning.
Complete HR data architecture visualization with interactive intranet portal: recruitment through offboarding, system integrations, and workforce analytics.
AI-powered voice workflows that deflect routine calls, accelerate triage, and hand off to agents with full context — boosting CSAT while slashing handle time.
Upload an MP3 (or WAV/M4A), auto-transcribe it, extract key entities, and generate a searchable, well-formatted PDF report with timestamps and highlights.
Full-featured audio player with multi-user stations, public/private visibility, real-time spectrum visualizer, playlists, and smart discovery—built for radio-style streaming at scale.
Full-featured music streaming platform with user authentication, album libraries, live radio streams, and smart search—delivering a Spotify-like experience with real-time playback controls.
Triage and rename PDFs by reading the first pages with an LLM to propose clean, consistent filenames; ideal for invoices, reports, and receipts.
Indexed 4M+ pages across PDFs & manuals; latency < 400ms; 32% fewer escalations.
Time-series models reduced unplanned downtime by 18% across fleet.
Top-1 accuracy 92% on 1,800 classes; mobile-first UI for field techs.
Real-time STT/LLM/TTS with agent handoff; cut hold time by 41%.
Low-latency speech and text translation across 80+ languages; < 200ms glass-to-glass; robust code-switching.
Android-friendly web page that opens apps via intent://, geo:, and custom schemes; floating YouTube/Maps; mic→ja-JP STT with auto-translate to English; safe-area tab bar.
PHP front-end for local OpenAI-compatible vLLM (Mistral-7B): mic capture via Web Speech, auto-detect EN/JA, auto-target or override, AJAX JSON responses, tunable temperature/host/model.
Web simulator and desktop tool to connect, scan ECUs, read/clear DTCs, stream live data, and save diagnostic sessions.
Online VIN decoder using NHTSA vPIC with a human-readable report, sample VINs, optional Wikipedia image, and raw JSON export.
Web-based FFmpeg scene change detector with adjustable threshold, optional thumbnails, and per-scene clip export.
Desktop tool to detect and read vehicle license plates from images or webcam using OpenCV + EasyOCR, with optional IP-based geotag overlay.