SkyDeck
The operating system through which the Indian Army conducts drone reconnaissance, fire services locate missing persons in disaster zones, and Indian Railways inspects critical infrastructure from the air. I worked on this for two years as part of a two-person design team, building the product from early wireframes through to the shipped platform used in the field today.
I set up the design system and interaction patterns for 10+ modules, with a focus on edge case design for high-stake scenarios. Live streaming, 3D photogrammetry, AI-powered analytics, geospatial comparison tools, mission planning, fleet management, and the admin console. Used across agriculture, telecom, construction, mining, defence and surveying. A subsidiary of Jio Platforms (Reliance Industries).
Live monitoring with persistent split-screen
The original design had a toggle between video and map. We started with the question: how do we give mission managers complete situational awareness without overwhelming them? During field testing with drone operators, we saw them constantly switching between the two views, losing spatial context every time. I worked with my senior designer to replace the toggle with a fixed split layout. Video on the left, map and telemetry on the right, always visible. That 1-2 second re-orientation delay we eliminated matters when Delhi Fire Services is searching a collapsed building from above.
3D tower inspection with automated detection
Telecom towers need regular inspection for rust, loose dishes, missing antennas. Traditionally someone climbs the tower. Now a drone orbits it, builds a 3D model, and AI counts the hardware and flags damage. I designed the Insights panel to sit next to the 3D viewport so the operator sees the evidence and the data at the same time. The AI confidence score is visible because operators need to know when to trust the machine and when to verify manually.
Precision agriculture with spatial data overlay
A colour gradient tells you something is stressed, but not by how much. Working with agronomists during research, I learned they needed precision, not just patterns. I added numeric NDVI values at grid intersections so they could make planting decisions from specific numbers, not just colour. The mission history timeline at the top lets them compare the same field across growing seasons. This was one of the clearest examples of adapting a data visualization pattern for a domain-specific need.
Construction progress through time-based comparison
Nov 2022 on the left. Aug 2023 on the right. One draggable divider. Project managers at Indian Railways and construction firms use this to verify progress against plans. Both sides have their own layer controls so you can toggle different data on each time slice independently. I designed both orthomosaics to load simultaneously, not sequentially, because the comparison IS the decision tool and a loading state in the middle breaks the thought.
Mission planning from approval to flight path
A mission moves through three states: scheduled, approved, in-progress. I designed the full end-to-end flow from the wireframe stage through high-fidelity prototypes, testing each step with mission managers. The pilot gets assigned, the drone and payload are selected, and every flight parameter feeds directly into the map. GSD determines altitude. Line spacing determines the scan pattern. Area determines battery count. Change one number and the flight polygon redraws in real time. This was the most complex screen I worked on because the data panel and the map had to behave as one connected system.
The hardest part of this project was not the screens you see above. It was designing for what happens when the system fails. Drone telemetry is not always stable. Signals drop. Batteries drain. No-fly zones appear mid-flight. I spent months working through these scenarios with my senior designer and the engineering team.
When a drone loses its connection, the icon on the map does not disappear. It stays at its last known position, visually ghosted, so the operator's mental model of the mission stays intact. When battery drops below threshold, the system does not just show a notification. It escalates the mission card in the dashboard, pulling the operator's attention before the situation becomes critical.
I also designed the annotation system by borrowing from YouTube's video timeline and adapting it for spatial data. Instead of just timestamping a comment, every annotation anchors to a GPS coordinate and a specific video frame. Post-mission, engineering teams do not have to scrub through hours of footage. They jump directly to the flagged location. That reduced review time by about 40%.
SkyDeck taught me how to make complex systems feel simple without removing complexity. The data is dense. The stakes are real. The users are working under pressure. Every design choice had to account for that context.
I learned that adapting consumer patterns to mission-critical contexts is harder than building from scratch. YouTube's timeline works for entertainment. Making it work for a drone operator who needs to tag a damaged power line at a specific GPS coordinate during a live flight is a different problem entirely. The pattern is familiar. The consequences of getting it wrong are not.
Working across five industries (agriculture, telecom, construction, mining, defence) showed me that the same interface can serve very different people if the information hierarchy is right. A farmer and a military operator both need to see location, status, and alerts. What changes is the priority, the urgency, and the cost of missing something. Designing for that range of stakes in one product was the most formative experience of my career so far.
This project changed how I think about design. Before SkyDeck, I thought good design meant getting the interface right. After two years of watching operators use the product in the field, I realized the real work is in what happens when the interface cannot help you. A drone loses signal. A battery drops mid-flight. A mission goes off-plan. The system I'm most proud of is not any screen you see above. It's the one that keeps the operator oriented when everything else is falling apart.