Senior Software Developer · UtopiaTech
I build the layer between hardware and humans — real-time IoT platforms in Angular, TypeScript and Node.js, where thousands of devices report in and the dashboard has to stay honest.
Mumbai, IN · Angular / Node.js / MQTT
EazyIoT — hardware-agnostic IoT platform
EazyIoT — dashboards & data visualization
React.js · internal tooling
Embedded systems · diagnostics
Most IoT platforms are built for one vendor's hardware. EazyIoT is built for whatever arrives next. Everything reaches the platform over MQTT — but what's inside the payload never agrees: one controller sends clean JSON telemetry, another emits binary frames that have to be decoded byte by byte, a third sits behind a gateway relaying register blocks from a panel that has never heard of the internet. Same platform, same dashboards, no rewrite.
Every reading a device sends crosses five layers before it becomes a number someone acts on. Each one is where a different class of bug lives.
One transport for every device. Topic structure carries identity and direction, so routing a new device family costs nothing at the edge.
Paired in/out parsers per device type, resolved at runtime. JSON payloads, binary frames and relayed register blocks go in — one flat, normalized shape comes out.
Device state and config in MongoDB, high-frequency telemetry in InfluxDB — separated so neither query pattern starves the other.
Scheduled cloud functions turning raw current and voltage into runtime, uptime, faults and outage windows.
Drag-and-drop dashboards, per-tenant layouts, charts that stay responsive with thousands of points on screen.
Adding a device family shouldn't mean touching the core. Each device type ships a key map and a parser pair, so a structured telemetry payload, a binary frame and a gateway-relayed register block all land in the same normalized document.
The hard half is the write path: turning a schedule a user drew in a browser into the exact byte layout a controller expects, then reconciling the acknowledgement that comes back. Config is a conversation, not a POST.
No device sends "this fixture was on 91% of last night." That gets inferred: measured current against rated wattage tells you how many lamps in a fixture are alive; relay state tells you which hours even counted as operational.
Most of the difficulty is time. Cron jobs run in UTC containers on IST infrastructure, outage windows cross midnight, and a day boundary computed in the wrong zone silently shifts every report by five and a half hours. The rules are resumable and day-wise by design so a restart never double-counts.
A handler that opened a fresh database client per incoming message worked perfectly in testing and exhausted the server's file descriptors in production. The fix was small — one pooled client, a raised container limit — but finding it meant reading the failure from the OS down, not the app up.
Deployments got the same treatment: scripted pull, build swap and service recreate, so shipping is boring on purpose.
St. Francis Institute of Technology, Mumbai · CGPA 9/10
For leadership and on-time delivery of key IoT product milestones at UtopiaTech.
Winner — UtopiaTech Hackathon and the Odoo Developer Challenge.
Open to interesting problems
Best reached by email — I answer quickly. Happy to talk IoT architecture, Angular performance, or anything that streams.