Killedar: a fort guide that works where the signal dies

Description and use

Killedar is an offline PWA for Sahyadri fort treks that uses Gemma and TabPFN to provide planning, on‑device Q&A, and narrated history without network.

Problem

AI trek assistants fail when network drops, leaving hikers without guidance at trailheads.

Outcome

Killedar enables offline planning, Q&A, and narration for Sahyadri fort treks, reducing reliance on network and improving user experience for early‑morning departures.

Workflow

• Plan: choose fort, trail, and last bus time; app predicts personal climb time using TabPFN and works backwards to give departure time.
• Ask: pose questions in English or Marathi; Gemma runs on-device via WebLLM and answers using only the fort’s notes.
• Listen: play pre‑generated ElevenLabs narration of the fort’s history in chosen language.
• Offline setup: tap "Use offline" on Wi‑Fi; app downloads predictions, Gemma model, and narration, then becomes usable in airplane mode.
• Use: during hike, rely on Ask and Listen without network.

Evidence

Killedar predicts your climb time from your past hikes and works backwards from that bus, so it tells you when to leave Pune and not just how long the climb takes.

Source

Ask: "Is there water on Padmavati Machi?" "Is the Chor Darwaja route OK for beginners?" Answered by Gemma running on the phone, grounded only in that fort's notes, in English or Marathi.

Source

Listen: the fort's history read aloud in English or Marathi, so you can look at the walls instead of the screen.

Source

On Wi‑Fi at home you tap Use offline once. Killedar saves the app, the fort notes, your predicted climb times, the narration and the Gemma model, shows the progress, and turns green when it's ready.

Source

TabPFN is a pre‑trained tabular foundation model, so "fitting" it on a few dozen rows is in‑context learning.

Source