The Kavaratti desalination engineer and the monsoon supply chain

His work is less visible. Underneath Kavaratti, in a concrete plant on the eastern edge of the island, Aravind manages two reverse-osmosis desalination units that together produce twelve thousand litres of drinking water per day — the entire freshwater supply for Kavaratti's permanent residents, roughly 12,000 people. He is not the only person there; he supervises three technicians and coordinates with a rotating engineer from the CPWD, who oversees the solar-panel array that powers the plant. But the logic of the system — when to run which unit, how to sequence the salt-water intake during monsoon tides, when the intake filters need cleaning, what the membrane pressure readings mean when they drift — that is Aravind's responsibility.

The plants have operated continuously since 2013, which is to say they have never stopped because they cannot. There is no other source of freshwater on an island where rainfall is concentrated into four months and groundwater is brackish. When the desalination plant operates, Kavaratti has water. When it does not, Kavaratti has a crisis.

What happened in July 2025 illustrated this with unusual clarity.

🗓️ The annual ritual

The Southwest monsoon in Lakshadweep runs from June through September. The rain is simultaneous and violent: eight inches can fall in a single night. The sea rises, the tide tables invert, the intake seawater temperature drops, and — most crucially — the shipping route from Kochi to Kavaratti becomes unpredictable. Ferries are cancelled sometimes weeks at a time. Container ships sit in port. Prices for anything that must be imported spike by 30 to 40 per cent, and availability becomes binary: either the freight arrived before the monsoon, or it arrives in October.

For the desalination plant, this seasonal wall has meant, for nine years, a single unvarying protocol: by the end of May, all anticipated spare parts for the June-to-September period must already be on the island. Filters, membrane cartridges, coupling gaskets, pressure-gauge diaphragms, electrical connectors — everything is ordered in the preceding February and March, when the procurement windows are open and shipping is certain. What is not on hand by June is simply not available.

This has never been a problem because Aravind has learned to anticipate failure with unusual accuracy. He reads the membrane differential pressure every morning, he knows the exact date a set of filters has been in service by their invoice numbers, he has a mental calendar of which components have a known failure interval. He discusses the history of the plant with the technicians each Monday morning. He keeps a spiral notebook in the control room where equipment serial numbers, replacement dates, and failure patterns are logged in neat handwriting. The notebook is less a regulation than a habit, the way another person might keep a diary of weather.

In February 2025, he submitted his spare-parts requisition to the Water Board office in Kochi. The list was conservative: two membrane cartridges (the RO membranes typically last 2.5 to 3 years and the current set was at the 2.8-year mark); four filter housings; six coupling gaskets; an electrical connector for the intake pump motor; a pressure-regulator diaphragm. Total cost: ₹3,84,000. Everything standard, everything with a known lead time. The order was approved in early March. By the first week of June, everything had arrived in sealed cardboard boxes stored in the cool room behind the control room.

"കടലിന്റെ കാലമതി എന്നൊരു വിഷയമുണ്ട് — കരയിലെ ആളുകൾ അത് മനസ്സിലാക്കില്ല."

— There is a season of the sea — people on land will never understand it.

By the end of July, he had used two of the four filter housings and one connector. The membrane cartridges were intact. The pressure-regulator diaphragm sat in its box. The protocol worked because Aravind's experience was precise.

What Aravind could not anticipate was the specific failure: on the morning of July 19th, at 6:47 AM, while he was walking to the plant from his flat, the secondary RO membrane ruptured. Not slowly; not with the gentle drift of differential pressure that meant a cartridge was aging. A complete, structural rupture — a pinhole breach that bloomed into a three-millimetre fissure in the membrane housing itself.

The plant went offline. In Lakshadweep, offline means offline. The primary unit was already handling the reduced flow during the monsoon tides. There was no backup. The reserve water in the overhead tanks would serve Kavaratti for approximately thirty-eight hours.

⚠️ What very nearly happened

Aravind received the call from the duty technician Vasant at 6:52 AM. He was still a hundred metres from the plant. He ran the last distance. The secondary unit control panel showed zero output pressure — a screen of red alarm indicators. The membrane housing itself was weeping — a thin ribbon of brine seeping from a hairline fracture in the stainless-steel casing. Replacement meant the entire membrane cartridge, which should have had another six months of service life ahead of it. He called the Water Board office in Kochi.

The officer on duty, a man named Muralidharan whom Aravind had known for four years, said: "This is July. Nothing arrives from the mainland until October."

Aravind did not immediately respond. He understood that Muralidharan was not being philosophical; he was stating fact. The monsoon had been heavy that year. The ferry schedules had already been compressed. There was no membrane cartridge in the Kochi warehouse. There was one in the CPWD depot in Cochin — Aravind learned this by noon, after a series of calls — but it was not a routine stock item, and getting it on the next available cargo ferry would require an emergency transportation exemption from the administration, which moved slowly, and even if approved, the ferry schedule was uncertain.

What very nearly happened was this: Kavaratti would operate on a single desalination unit at 80 per cent capacity for the next 100 days, until October. A single unit running at near-maximum pressure accumulates damage. The intake system would experience stress that aged it faster. A backup of salt water in the secondary line, if not manually purged each morning, could corrode the coupling gaskets. The overhead tanks would be emptied and refilled three times daily instead of twice. The hydrogen sulphide smell from the treated water would become noticeable.

More consequentially, if the primary unit failed — and a component failing a year after installation was not theoretically impossible — the island would be without water while the ferry schedules normalised.

Aravind sat in the control room that evening, after manually isolating the secondary unit and verifying the primary unit's output pressures. The pressure readings were good. The intake temperature was dropping with the monsoon cooling. At current consumption, the primary unit could sustain the load for four months. But "could sustain" and "should sustain" were different assertions.

He typed a formal incident report to Muralidharan that night, noting not just the failure but the supply-chain contingency. The reply came in the morning: a digital shrug. Emergency procurement from the Cochin depot was pending administration approval. Approval typically took two to three weeks. The ferry schedule was unpredictable. Plan for October.

He called Asha from the plant. He said: "The secondary unit is offline. It will be offline until October." She heard the tone and said nothing. She understood that he would work through this, methodically, the way he worked through everything, and that he would not leave it to chance.

🌗 What changed

In the last week of August, three weeks into the monsoon, Aravind was sitting in the control room on a Thursday afternoon, logging the daily readings in his notebook, when Mira called. She said her school had gotten a new tablet for the staff office and she had installed something on it that she wanted to show him. He said he was busy. She said: "No, abba, it is a thing that helps with things. Install it." He heard the certainty in her teenage voice and agreed to look.

That evening, he took the old Samsung tablet from the flat — a hand-me-down from Asha's hospital that was no longer quick enough for medical files — and opened the application. Mira had set the language to Malayalam. There was a prompt asking for his problem statement.

He typed slowly: "കരണ്ടത്തെ വെള്ളത്തിനെ കുറിച്ച് ഞാനെ ആലോചിക്കുന്നത്. എനിക്കെ ഓരോ വർഷവും ജൂണ്ട് മുമ്പ് സാധനങ്ങൾ വാങ്ങേണ്ടി വരുന്നത് കാരണം അത് വിതരണ കഠിനമാണ്. എന്നാൽ എനിക്കെ ഭാഗങ്ങളുടെ ആയുസ്സ് എത്ര നാൾ വാണെന്ന് കൃത്യമായി അറിയാൻ പോലും ബുദ്ധിയിലെ ഒരു വിധിയിലും ഞാൻ കാണാത്തത് ഉണ്ടെ."

(I think about Kavaratti's water. Every year I must buy things before June because the supply is difficult. But I do not even have a proper way to know exactly when parts will fail.)

The agent responded in Malayalam:

"അരവിന്ദ് സാർ, അത് ഒരു പ്രാധാനിക പ്രശ്നമാണ്. നിങ്ങളുടെ എത്ര വർഷത്തെ നിരീക്ഷണം സമ്പർണ്ണമാണ് — എത്ര വർഷം പരിപാലന സായ്കലുമുണ്ട്? എന്നാൽ ഇത് കേവലം നിങ്ങളുടെ മനസ്സിൽ സ്റ്റോർ ചെയ്യുന്നതാണ്. കരണ്ടത്തെ വെള്ളത്തിനെ കുറിച്ച് സർവ്വതത്വം മാറ്റുവാനായി, നിങ്ങൾ ഓരോ ഭാഗങ്ങളുടെ പരിപാലന ചരിത്രം എനിക്കെ നൽകിയാൽ, ഞാൻ പരാജയത്തെ പുർവ്വാനുമാനം ചെയ്യാൻ കഴിയും — എത്ര പ്രത്യാശിതമായി."

(Aravind sir, this is an important problem. Your nine years of observation is complete — you have maintenance cycles. But this is stored only in your mind. To make Kavaratti's water permanent, if you give me each component's maintenance history, I can predict failure — exactly how many days ahead.)

The conversation took an hour. Aravind explained the secondary RO membrane that had ruptured on July 19th. He gave the installation date: January 2023. He gave the membrane cartridge model number and the technical specification: nominal lifespan 2.5 to 3 years at standard operating conditions. He explained that monsoon tides meant the intake temperature dropped by 4 to 6 degrees, which theoretically extended membrane life but increased salt-concentration variance. He explained the coupling gaskets, which had a known failure interval of eighteen months under high-pressure conditions.

The agent said: "നിങ്ങൾ ഇത് നിരീക്ഷണത്തെ കഠിനമായി കണ്ടെത്തിയിരിക്കുന്നതാണ് — കിന്തു ഇത് ഏകാധികാരത്തിന്റെ ഒരു നിരീക്ഷണമാണ്, പരിപാലന തയ്യാരിയുടെ അല്ല। നിങ്ങളുടെ നിരീക്ഷണത്തെയും സ്ഥാനങ്ങളെയും ഉപയോഗിച്ച്, ഞാൻ സാധാരണ തയ്യാരി നിയന്ത്രണാധികാരകഷ്ണ നിർമ്മാണം ചെയ്യാൻ കഴിയും — ഇത് നിങ്ങൾ കരണ്ടത്തെ സ്ഥാനത്ത് മെയ്, ജൂൺ ഓരോ വര്ഷവും നടത്തുകയാണ്."

(You have observed this harshly — but this is observation of incidents, not preparation for maintenance. Using your observation and locations, I can build you a maintenance schedule of what you normally control — this is what you must do in Kavaratti every May and June each year.)

Over the next week, they built a component inventory spreadsheet, a failure-history timeline, and a procurement calendar. The agent created a prediction: based on 2023 installation date and monsoon-adjusted lifespan estimates, the first coupling gasket would fail between April and June 2026. The secondary membrane, if replaced now (in August 2025), would reach end-of-life between October 2027 and January 2028. The primary-unit intake filters, currently installed in March 2025, would need replacement by December 2025 — well before monsoon.

But the structure was the radical shift: the agent produced a table with three columns — Component, Installation Date, Predicted Failure Range — and then a fourth column: Order Deadline (Latest). Every component now had not just a failure date but a procurement deadline with a margin.

1. ⚠️

   July 19, 2025 — Membrane rupture

   Secondary RO membrane failed completely. No spare available. Island on single-unit operation for 100 days, with no backup if the primary unit failed.

2. 🗂️

   August 2025 — Component inventory audit

   Aravind transcribed nine years of maintenance notebooks into a structured spreadsheet with installation dates, failure patterns, and monsoon-adjusted lifespan data.

3. 📋

   September 2025 — Predictive maintenance schedule

   Agent generated a 24-month procurement calendar with failure predictions and Order Deadline columns. Components ranked by criticality and supply vulnerability.

4. ₹

   May 2026 — Advanced procurement cycle

   Aravind submitted orders in February 2026, two months ahead of the May deadline. All monsoon-season spares secured before shipping shut down. Zero emergency procurement in the 2026 monsoon.

He showed the spreadsheet to Muralidharan in Kochi. Muralidharan said: "This is good. We can work with this. But you understand that the administration still must approve each procurement order." Aravind said he understood. What he was telling Muralidharan was simpler: I can now tell you, with five months' warning, when I need what. You can tell the administration. The administration can approve on a schedule, not in crisis.

The breakthrough was not technical. It was administrative discipline applied to pattern recognition. Aravind had always known the patterns. He had written them in his notebook. The shift was making them visible — to the Water Board, to Muralidharan, to the island administration's procurement cycles. The agent had not replaced his judgment; it had made his judgment legible.

By October 2025, when the ferry schedules normalised, the secondary membrane cartridge arrived. The emergency procurement exemption was never needed. Aravind installed it on a Monday morning, verified the pressure readings, and logged the installation in his spreadsheet. The primary unit had carried the island safely for 100 days. But the shape of the risk had shifted.

🧭 Why we built it

In Lakshadweep, there are approximately twenty-three engineers across the UT Working for the UT Water Board, CPWD, and Energy Department. Most serve on rotational postings: two to three years on the island, then mainland transfer. Aravind is an exception — nine years, no transfer posted, no indication he is asking for one. He has become native to the problem in a way that mainlanders cannot, because he has seen nine monsoons in a row, not one or two across separate postings.

But what makes him an exception also constrains the system. Aravind's knowledge is his alone. The moment he is transferred — or the island loses power and all the data on the control-room computer is corrupted, or he becomes unable to work — the knowledge defaults to the maintenance manual, which says "replace every 2.5 to 3 years" without accounting for monsoon temperature variance, without knowing that coupling gaskets fail at eighteen months under local high-pressure conditions, without the specificity that only observation can create.

The problem is not unique to Kavaratti. Across the UT, water systems, power grids, and coastal structures are maintained by a tiny, rotational engineering corps. When they leave, their nine years of hand-written observations leave with them. The system reverts to the manual.

What it does

• 🔍Reads Aravind's maintenance history, identifies patterns in component failures, and generates a calendar of predicted failure ranges with monsoon-adjusted parameters.
• 🗂️Organizes nine years of hand-written notes into a structured spreadsheet with installation dates, failure dates, and procurement deadlines — all readable by the next engineer.
• 📞Flags upstream bottlenecks: shows the Kochi Water Board office exactly when to submit procurement orders to align with administration approval cycles.

What it does not do

• 🔒Never makes a decision about component replacement — Aravind remains the sole authority on when to order, how many units, and what model specifications.
• 💳Never submits a procurement order or interacts with the administration — the schedule is a recommendation, the decision is his.
• ✅Never replaces observation with prediction. The model lifespan is a guide. Aravind's daily pressure readings, his Monday notes, his understanding of the monsoon cycle — those remain his.

The second reason is more subtle. Muralidharan in Kochi coordinates procurement for five island units: Lakshadweep, Andaman & Nicobar, Puducherry, Daman & Diu, and Chandigarh. He has a central budget. He has administrative cycles. He has ports and shipping schedules. When an engineer on a remote island says, in July, "I need a spare part in September," the request lands in his inbox as an emergency, with all the friction that entails. When the same engineer says, in February, "Here is a calendar. In May I will need this. In September I will need this. In November I will need this," the request integrates into his planning. It integrates into the administration's budget cycle. It becomes a routine.

Aravind's spreadsheet will not be visible to the next engineer unless someone preserves it. But it is now visible to Muralidharan, and that is enough.

🌱 What we hope happens

In April 2026, Mira asked her father whether the desalination plant had "stopped breaking things." Aravind said it had not stopped breaking — things break. What had changed was knowing, five months ahead, that a thing would break, and being able to tell Kochi and the administration with precision rather than panic.

The product is free. We will keep it free for Aravind — the engineer in Kavaratti who has transcribed nine years of hand-written observation into a spreadsheet so that the next engineer, whenever that engineer arrives, will inherit not just a plant but the knowledge of how that plant behaves under monsoon tides, what ages it, what does not, and when the ordering season actually is. It remains his responsibility. It remains his decision. But it is no longer hidden in a notebook.

The UT Water Board estimates that maintenance costs across five island units could be reduced by 8 to 12 per cent if supply-chain procurement were planned rather than reactive. A single emergency overseas shipping exemption costs ₹2,50,000 and takes four weeks to process. If Aravind's model is applied to all five units, the administration estimates six to eight fewer emergency shipments per year across the UT network. That is administrative friction reduced, cost saved, and infrastructure more reliably supplied during the season when replacement parts matter most.

What we hope happens is quiet: that the next time a membrane ruptures on a remote island in July, someone does not tell the engineer that nothing can be done until October. That procurement cycles integrate observation the way a well-run workshop always has — with the person who knows the machine sitting at the table when the order is discussed. That the monsoon comes, Aravind runs the plant, and the island has water. That the knowledge is written down and legible, so that if he leaves, if the computer fails, if the administration changes, the observation survives.

He called us in May 2026 to say that he had installed the predicted maintenance schedule on the Kochi Water Board office tablet, and Muralidharan had started using it for the other island units. He said: "The monsoon is coming. The orders have been submitted. I have no emergency procurement pending. That has not happened in nine years."

The product exists, and it is free. For the engineer who knows his plant well enough to write down what works. For the administrator who knows that the engineer's knowledge is worth more than the process that usually captures it. For the island that needs water all year, not just until June.