Patrick Collison keeps a page called Fast — a list of people who did ambitious things quickly. When it gets shared, the examples that travel are the big ones: the Empire State Building, Apollo 8, the Eiffel Tower. What gets lost is the other kind of example, the one that is closer to what most of us actually do: a handful of people in one Indian city, no press, no ribbon, working on something nobody had asked them to build.
I started collecting these for myself. Two things kept happening. First, almost every time I told someone about one of them, they hadn't heard it. Second — and this is the reason I'm writing it down — the lesson in each case was never "work harder." It was something specific and transferable: a constraint that removed half the design space, a person who refused to wait for an institution, a network of borrowed instruments, a failure that arrived in distribution rather than in the design.
So this is a list, in the spirit of Collison's, of engineering done on Indian soil by teams small enough to fit in a room — ten from before 2010, four from after. Two of them failed. I've kept those in, because a list of successes teaches you what to copy, and a list with failures teaches you what to check.
The list
Fourteen at a glance
| # | What | Team | Years |
|---|---|---|---|
| 1 | India's own telephone exchange | avg age 23 | 1984–86 |
| 2 | PARAM 8000 | one team | 1988–91 |
| 3 | The Trombay plutonium plant | first 3 trainee batches | 1959–64 |
| 4 | The TTK-Chitra heart valve | 3 people, then 5 | 1980s–90 |
| 5 | Aurolab's $2 lens | a hospital's trust | 1992– |
| 6 | The Jaipur foot | a surgeon + a craftsman | 1968– |
| 7 | Bhoomi land records | ~4 engineers + 1 officer | 1998–2003 |
| 8 | The Simputer ✕ | 7 people | 1998–2001 |
| 9 | Mitticool | one potter | 2001– |
| 10 | chotuKool | 2 people in a conglomerate | 2007–13 |
| 11 | TeamIndus ✕ | 100+ engineers | 2010–18 |
| 12 | Skyroot's Vikram-S | 200 employees | 2018–22 |
| 13 | Agnikul's printed engine | IIT-M incubated | 2017–24 |
| 14 | Chakr's soot-to-ink | 3 IIT-D graduates | 2016– |
Before 2010
1. India's own telephone exchange
C-DOT was set up in August 1984 with a promise that sounds like a dare: a family of digital telephone exchanges in 36 months for ₹36 crore. The team was hired fresh out of college — average age twenty-three — on the theory that nobody had yet taught them what was impossible. Their first product, a 128-line rural exchange, went live at Kittur in Karnataka in July 1986. It had to run without air conditioning in Indian heat, dust and monsoon humidity, which the imported switches could not: so the engineers deliberately ran the processors slower, spread the components out, and shaped the cabinet to chimney heat away. Bulkier, slower, one-tenth the cost of imports.
The part I keep coming back to: they announced "an RAX a day", nothing happened for a month, a newspaper called it A Hoax a Day, the engineers took it personally, and within a month they were at two a day. Eventually ten. By 1993, twenty-five.
The constraint was the specification. Because the exchange had to survive a village with no air conditioning and an unreliable grid, it became cheap enough to put in every village. Designing for the worst conditions and designing for the poorest customer turn out to be the same engineering problem.
Sources: C-DOT's timeline · Pitroda on the design decisions
2. PARAM 8000
When the US refused to sell India a Cray supercomputer — the fear being that it would end up in weapons design — the answer from the Prime Minister's technology adviser was: we'll build one. C-DAC was created in 1988 with a three-year runway and ₹30 crore, roughly the price of the Cray that had just been denied. Vijay Bhatkar's team in Pune worked on parallel processing and assembled the machine from 64 transputer nodes with a reconfigurable network. A prototype was benchmarked at the 1990 Zurich supercomputing show and placed second, behind only a US machine. PARAM 8000 shipped in August 1991 at about $350,000, a fraction of a Cray, and was later exported to Germany, the UK and Russia.
They did not try to reproduce the machine they were denied. They bought the one part that could not be made in India, designed around its weakness with a different architecture, and competed on a different axis. Refusing to copy the thing you were refused is a strategy, not a compromise.
Sources: How C-DAC was set up · PARAM 8000 specifications
3. The Trombay plutonium plant
India decided in December 1958 to build a plant to reprocess spent fuel from its own reactor. Design work ran from January 1959 to January 1961; the plant was commissioned in 1964, inside its schedule and its cost, designed and built entirely in-house by engineers from the first three batches of the BARC training school — Homi Sethna as project engineer, N. Srinivasan as design engineer — with about ₹50 lakh of equipment fabricated in a workshop on the site. Total: ₹3.6 crore, less than half the estimated cost of a comparable-capacity plant then being built at Mol in Belgium by a consortium of twelve European governments and twenty-two chemical firms. India became the fifth country with reprocessing capability.
The team was not experienced, it was recently trained and given the whole problem. There had been almost no pilot plant. The bet was that a small group which had learned the physics together, and owned design, fabrication and commissioning, would beat a consortium with better credentials and a worse feedback loop.
Sources: IGCAR's history of Indian reprocessing · Contemporary US consular dispatch
4. The TTK-Chitra heart valve
In the early 1980s, a hospital in Thiruvananthapuram that treated poor patients free could not afford imported heart valves, so its waiting list simply grew. A cardiac surgeon (M. S. Valiathan), a biomedical engineer (G. S. Bhuvaneshwar) and a veterinary surgeon (Arthur Vijayan Lal) decided to make one — a decision met, in Valiathan's words, with "disbelief verging on derision". They set two conditions before starting: the valve would meet ISO standards, and it would be affordable to an Indian patient. Together, those two constraints are what made the project hard.
Model one failed: a weld fractured, and the National Aerospace Laboratory found oxygen contamination had embrittled it. Model two failed: the titanium housing wore against a sapphire disc until the disc could escape. Model three passed every test and then a sheep dropped dead from a fractured sapphire disc. Model four — a cobalt-chrome housing with an ultra-high-molecular-weight polyethylene disc — took under six months, passed everything, and went into a patient on 6 December 1990. That patient was still well decades later. The valve has been implanted in more than 100,000 people and pushed the price of imported valves down by roughly half.
How they did it with no laboratory is the part worth stealing: ISRO's pressure-transducer facility did the electron-beam welding, a textile research association in Coimbatore knitted the sewing ring, the National Chemical Laboratory developed the polyethylene, and DuPont sent 20 kg of polymer free when asked for a pro-forma invoice.
A hard constraint stated at the start ("ISO-compliant and affordable") is worth more than a large budget, because it eliminates the design space where you would otherwise waste five years. And the network of borrowed instruments was not a workaround — it was the method.
Sources: Valiathan's account of the four models · DST on the valve's scale
5. Aurolab's two-dollar lens
Before 1992, the consumables for a cataract operation cost over $100, which put the surgery out of reach of most Indians. Aravind Eye Care, which began in 1976 as an eleven-bed rented house in Madurai, concluded it could not solve blindness by waiting for prices to fall, and set up a non-profit trust to manufacture intraocular lenses itself. The engineer who led it, P. Balakrishnan, left a research post at the University of Wisconsin and came home. They bought process know-how from a small Florida lens maker, built the equipment to suit Indian labour economics rather than the other way round, and trained nurses from the hospital as production staff. The first lenses went out at $10; rigid PMMA lenses now sell for about $2–3. Aurolab has made over 20 million lenses, exports to roughly 140 countries, and produces about a tenth of the world's supply.
There is a second half usually left out. By 2001 at least eight Indian firms were making cheap PMMA lenses, and a masked bench evaluation of ten lenses from each found that only one manufacturer met every optical and mechanical specification. Several had critical defects, including unpolymerised monomer. Cheap turned out to be the easy part.
The price breakthrough was an organisational achievement, not a technical one — a non-profit with the same equipment and the same standards chose a different pricing objective. The copycats prove the corollary: if the quality system does not travel with the price, the cheap version is not the same product.
Sources: American Academy of Ophthalmology on Aurolab · Aravind's 20-year review
6. The Jaipur foot
P. K. Sethi, an orthopaedic surgeon in Jaipur who had never trained as one, noticed that the artificial limbs his patients were fitted with assumed a patient who wears shoes, sits on chairs and walks on pavements. His patients took their shoes off indoors, sat cross-legged, squatted, and worked barefoot in wet fields. So in 1968 he sat down with Ram Chandra Sharma, a craftsman with no formal education, and they wrote seven requirements before designing anything: no shoe needed, waterproof, enough ankle flex to squat, rotation for cross-legged sitting, inversion and eversion for uneven ground, inexpensive, and made of locally available materials. Two years of experiments followed. The breakthrough came from a punctured bicycle tube: cast it in rubber. The first version shredded, so they hinged rubber to a wooden limb, wrapped it lighter, and vulcanised it.
The published paper is a lesson in empirical engineering. Field failures cracked around a wooden block, so they bound the foot in cord lining the way tyres are reinforced. The block split along the grain, so they laminated it like plywood. And when skiagrams showed the sponge-rubber layers had separated during vulcanisation, they found the separation improved mobility, because the layers could slide on each other. In 1978 the material cost was about £2. BMVSS, set up in 1975, has since fitted roughly 1.4 million limbs, free, in about 45 minutes each.
The specification came from watching users, not from the standard. Sethi's advantage was that he had not been trained in orthopaedics — he had no inherited design to defend — and Sharma's was that he knew what rubber and wood could do. The pair of them beat better-funded prosthetics because they wrote the requirements down first.
Sources: Sethi's 1978 design paper · UNDP on the collaboration
7. Bhoomi
In 1998 a Karnataka civil servant with a computer-science background, Rajeev Chawla, was given the job of computerising the state's land records. Under the manual system, 9,000 village accountants held the only copies; a farmer needing a record of tenancy waited three to thirty days and paid a bribe of $2–40, more if he wanted the details written vaguely. The project digitised 20 million records covering 6.7 million farmers — about a billion individual data entries, done by roughly 150 staff at private firms over 18 months — printed every record and had village accountants verify it against the manual register, then changed the law so the manual records ceased to be valid. Records were issued from kiosks in 177 taluk offices for ₹15. The software itself was written by a team of four at the National Informatics Centre, with a technical director spending perhaps a tenth of his time on it.
It was not frictionless, and the criticism is fair: the kiosks were often a day's travel away, so the better-off gained most, and the model has never replicated as cleanly elsewhere.
The technology was the small part. The leverage came from three non-technical decisions — invalidate the old records by law, staff the kiosks with new recruits who had never enjoyed the old power, and verify every digitised record against its paper original before switching over. Most e-governance projects fail at those three points, not in the code.
Sources: Princeton's Innovations for Successful Societies case · Chawla's account for the World Bank
8. The Simputer ✕
In November 1998 a group at Bangalore IT.com wrote the Bangalore Declaration on IT for developing countries. Seven people — four from IISc's computer science department, three from a local software firm — spent about fifty hours arguing about what a device for the unreached would have to be, then built it: a Linux handheld running on three AAA batteries, with text-to-speech in Indian languages so a non-literate user could operate it, and a smartcard slot so one device could be shared by a village, each person carrying their data on a card. Target price ₹9,000 at volume, against about ₹33,000 for the nearest comparable imported device. The development, worth roughly $5 million, was donated. The New York Times called it the most significant innovation in computing in 2001.
It then sold almost nothing. There was no hardware supply chain in India, no venture capital for hardware, and although the device was demonstrated repeatedly to central and state governments, no order came. The people who most needed it could not afford ₹9,000, and the institutions that could afford it preferred to buy from vendors they already knew.
The engineering was not the binding constraint. A product whose customer is a village and whose payer is a government needs a procurement story before it needs a prototype. Swami Manohar's own summary is the line to remember — new technologies have to be built on the assumption that they will fail, and funded accordingly.
Sources: Frontline's report from the launch · Economic Times on why it flopped
9. Mitticool
After the 2001 Gujarat earthquake, a newspaper photograph of shattered earthen pots carried the line garibi ka fridge tut gaya — the poor man's fridge has broken. Mansukhbhai Prajapati, a potter's son from Morbi who had failed class ten and worked odd jobs, took it literally. He spent five years building a refrigerator out of clay: water in an upper chamber drips down the walls, evaporates, and pulls heat out of the chamber below, holding it five to eight degrees below ambient. No electricity, no moving parts, no recurring cost, about ₹3,000. He worked alone; the National Innovation Foundation later helped with a patent, testing and a first advance.
He did not invent a technology, he re-derived one — evaporative cooling, which every village in India already knew — and then did the hard part, which was turning a household habit into a product with a shape, a shelf, a tap and a price. Familiar physics plus unfamiliar packaging is an underrated form of innovation.
Sources: NIF's technical description · NIF on how the idea started
10. chotuKool
Godrej set out in 2007 to build a refrigerator for the roughly 80% of Indians without one. Gopalan Sunderraman began alone and then had exactly one colleague. Their first instinct — a smaller, cheaper compressor fridge — failed on the arithmetic of ownership: purchase price, electricity, servicing and space all added up to too much for the customer they were aiming at. Field research in villages then changed the design three times. Many households that owned fridges kept them unplugged for fear of power surges, so constant mains supply was out. Families moved for work, so portability mattered more than capacity. And people already kept food cool in a sealed container in a water bath, so modest cooling was acceptable. The result used Peltier thermoelectric cooling — the technology that cools laptops — with 20 components and no moving parts, at about ₹3,500.
The pilot then taught them things the lab could not. The cooling was not good enough, so it had to improve by 35%. And buyers did not tell their neighbours about it: aspiration was growing faster than income, and a chotuKool advertised the fact that you could not afford a fridge. The customers who did get excited were not households at all but small shops, which could sell cold drinks at a higher margin — a market the team had not been looking for.
The product was designed three times — once in the office, once by the field research, once by the pilot — and only the third version was right. If you cannot afford three iterations, you cannot afford to serve a customer you have never met.
Sources: Livemint on the design and price · Harvard Business School case
After 2010
11. TeamIndus ✕
Six people registered a company in 2010 to win the Google Lunar XPRIZE: land a privately funded spacecraft on the Moon, drive 500 metres, send back video, all before a deadline. TeamIndus grew to over a hundred young engineers in Bengaluru, with former ISRO hands advising and Nandan Nilekani as the largest early investor. They built a lander and a rover — the rover named ECA, Ek Choti Si Asha, a small hope — won a $1 million milestone prize, and in 2016 signed a launch contract with ISRO's Antrix for a PSLV. The competition ended in January 2018 with no winner: no team could raise the money. TeamIndus had raised about half of the ₹400–450 crore the mission needed, and the launch contract was cancelled.
A prize is not a customer. The team built real lunar hardware on a fraction of what a national programme would spend — that part worked. What never existed was a revenue model that could fund the gap between milestone prizes and a launch, and the hardware could not be sold to anyone until it had flown. Fund the business model in the same sprint as the engineering.
Sources: The Hindu on why it stopped · Indian Express on the prize ending
12. Skyroot's Vikram-S
Two former ISRO engineers, Pawan Kumar Chandana and Naga Bharath Daka, founded Skyroot in June 2018 in Hyderabad. Vikram-S, a single-stage solid-fuel suborbital rocket, took about two years to develop and was built with carbon-composite structures and 3D-printed components. On 18 November 2022 it lifted off from Sriharikota, hit 89.5 km — past the 80 km target — and splashed into the Bay of Bengal about five minutes later, carrying three customer payloads. It was the first launch of a rocket built by a private Indian company, authorised by IN-SPACe. The company had 200 employees and ₹526 crore raised at that point, and its own engineers describe the rocket as a technology demonstrator for the orbital Vikram-1.
They flew something small and real before promising something large. A suborbital demonstrator is a strange product to sell — it earns little and proves a lot — and it converted a decade of institutional know-how into a company in four years. When you leave a large organisation to build, the first deliverable should be a demonstration, not a roadmap.
Sources: ISRO on Mission Prarambh · The Hindu on the launch
13. Agnikul's printed engine
Agnikul Cosmos started in 2017 out of the National Centre for Combustion R&D at IIT Madras. In February 2021 the team fired Agnilet, a semi-cryogenic upper-stage engine printed as a single piece of hardware in one run of a 3D printer, with zero assembled parts — injectors, cooling channels and igniter all designed into one part. "We don't think anyone in the world has ever pushed 3D printing of a rocket engine to this extent," said co-founder Srinath Ravichandran at the time. In May 2024, after four unsuccessful attempts, Agnibaan SOrTeD flew from Sriharikota — the first Indian rocket with a fully 3D-printed engine, the country's first semi-cryogenic-powered launch, and the first from a private launchpad in India.
Printing the engine in one piece was not a manufacturing stunt; it deleted the assembly step, and with it a whole class of failure modes and a supply chain of hundreds of parts. When a part count falls from hundreds to one, what you have removed is not cost — it is coordination. Four failed attempts before a first flight is also the honest number to quote when someone asks how long this takes.
Sources: The Hindu on Agnilet · PIB on the SOrTeD flight
14. Chakr's soot-to-ink
The idea came from a roadside sugarcane-juice stall in Gurgaon: the crusher ran on a diesel generator, and its exhaust pipe had been diverted at a wall, which had turned black. If that carbon can colour a wall, it can be ink. Arpit Dhupar, Kushagra Srivastava and Prateek Sachan — IIT Delhi graduates, all in their early twenties — built Chakr Shield, a retrofit device that fits a diesel generator's exhaust, cools the gases so the particles coagulate and can be trapped, and captures 70–90% of the particulate matter. The soot is then washed out with a solvent, stripped of heavy metals and carcinogens, and turned into pigment: about 34 litres of ink per generator per month, one litre printing 30–40 T-shirts. Mentors initially called the idea idiotic.
They solved a pollution problem by finding a customer for the pollutant. The engineering was clever, but the unlock was reframing a waste stream as a raw material — which is what made the capture economically survivable instead of a permanent subsidy. Look for the version of your problem where the by-product pays for the fix.
Sources: The Hindu on how Chakr Shield works · YourStory on the founding
Patterns
What recurs across all fourteen
The constraint is the design brief. Every one of these had a sentence that removed most of the design space before work began: no air conditioning and no reliable grid, no Cray, ISO-compliant and affordable, no shoe, no chair, no pavement, no electricity, no mains supply, no supplier of rocket engines. Teams with vague briefs and generous budgets did not do better; they mostly did not ship.
Small is a mechanism, not a mood. Three people for a heart valve. One officer plus four programmers. One executive plus one colleague. Three founders at a juice stall. The size is what lets a decision be made and reversed in a week. Where the technology was fine and it still stalled — the Simputer, TeamIndus — the ownership of the outcome was diffuse, and the missing function was never engineering.
Borrowed instruments beat built institutions. ISRO's welders, a textile lab's knitting machines, an aerospace lab's failure analysis, a chemical company's free polymer, a four-person software cell inside a national agency, a shed on the plant site that fabricated ₹50 lakh of equipment. None of them waited for a centre of excellence to be sanctioned.
Serendipity only pays if you are measuring. The Jaipur foot improved because a skiagram showed the rubber layers had separated and someone asked why that might help. The Chitra team found its winning disc material only after three failures had taught them exactly which property to optimise. Neither group was lucky in any useful sense; both were taking data.
Failure is usually distribution, not engineering. The Simputer worked and did not sell. chotuKool worked and sold to the wrong customer first. TeamIndus built hardware nobody could buy. Aurolab's imitators made cheap lenses and got the quality wrong. The last mile of an Indian product is procurement, channels and trust — and that is where the engineering reputation of the country is actually lost.
None of these people waited for permission, a budget cycle, or a hashtag. That is the whole list.
If you know one of these stories from the inside — or a better one that never got written down — I'd like to hear it. The list is long and my reading of it is shallow.
Sources & method