Reported
Harvesting at night
Buds are picked after dark so they are still closed when packed. Pickers work by headlamp in dense, waist-high beds.
Vaftsy AI & Robotics · Engineering-led innovation
Robots for the night harvest.
Bringing together engineering, artificial intelligence and robotics to solve real-world agricultural challenges. We are developing a machine designed to pick nishigandha (tuberose) buds at night, at the right stage, without damaging them.
Status: concept and research stage · no robot built yet
The problem
Agriculture has problems that cannot be solved by software alone.
Nishigandha is harvested at night as closed buds that are just about to open. Picked, packed and trucked for 4–5 hours, they open as they reach the market. The whole system depends on people picking the right buds, by hand, in the dark.
Grower photographs. A single spike carries buds at several stages, so each one has to be judged on its own.
Reported
Limits of manual picking
Growers report 11–12 kg per picker in a 4-hour night shift. Night labour is hard to find, and output falls as pickers tire.
Reported
Delicate buds
Growers estimate that about 10% of picked flowers are written off after handling damage, and rough picking harms the buds still on the spike.
Reported
4–5 hours to market
The journey from farm to market sets the clock. A bud is only right if it opens in that window, not before and not after.
Why picking each bud at the right stage matters
The same night's harvest, followed from the field to the market for buds picked at three different stages.
Bud picked
NightPicking
Before dawnPacking
4–5 hoursTransport
MorningAt market
Too earlyGreen, tight bud
Closed and green
Still closed
Does not open
Unopened, low value
Right stageWhite, plump bud
Closed, ready
Closed, packs well
Begins to open
Fresh and open
Too lateTip already splitting
Already opening
Open, easily bruised
Damaged in transit
Past its best
Simplified illustration based on grower descriptions. The AI task is to predict, for every bud, when it will open, and pick only those that will open during transport.
Why this needs both engineering and AI. A vision model can judge which buds are ready. Only a well-engineered machine can drive the bed at night, reach that bud and store it without bruising it or the buds around it.
Sources and how we label figures
Reportedstated by growers EstimatedVaftsy calculation Targetdesign goal, not yet achieved Nothing on this site is labelled Validated yet: no robot has been built or field-tested.
- 11–12 kg per picker per 4-hour shiftReportedVaftsy conversations with nishigandha growers near Pune, 2026. A picker's typical haul for one 4-hour night shift, as stated by growers. Not weighed or audited by Vaftsy.
- About 10% written offReportedGrowers' own estimate of the share of picked flowers discarded because of handling damage. Not independently measured.
- 4–5 hours to marketReportedTypical farm-to-market transport time described by the same growers.
- ~23 kg per robot per 4-hour shiftTargetDesign target for one robot over the same 4-hour shift as a picker. To be measured in field trials.
- Below 4% damageTargetProposed measure: the share of picked buds, by weight, rejected at grading for bruising or breakage, recorded during field trials against hand-picked buds from the same beds.
- ₹2.5 lakh per robotTargetTarget manufacturing cost for the first commercial version. Not a selling price or quotation.
The technology
Four capabilities, one machine.
Select a capability to see what it does. Every part is at concept stage today.
Concept visualisation · straddle chassis
Robotic mobility
Development stage
- Concept
- Designed
- Simulated
- Tested
How the four work together, bud by bud
- MobilityAdvance along the bedWheels stay in the paths
- VisionFind and grade budsStage and opening time
- ControlDecide what to pickOnly buds due to open in transit
- ManipulationPick and storeSoft grip, force-limited
- ControlLog and move onRepeat for the next spike
Engineering is our foundation
Built on engineering. Extended through intelligence.
Vaftsy AI & Robotics is part of Vaftsy, an engineering company practising since 2001. Our in-house CAE, mechanical, electrical and civil teams let us design and validate the complete robotic platform under one engineering organisation.
2001Engineering practice since 2001, when the civil wing was founded
3Established wings: civil (2001), electrical (2004), CAE (2007)
50+Technical and research papers authored by Vaftsy CAE engineers
Vaftsy Engineering is our parent company: integrated systems, CAE, civil and electrical engineering services.
Explore Vaftsy Engineering- Civil · 2001
Mechanical design and structural engineering
Chassis, frames and mechanisms designed for field loads.
- CAE · 2007
Finite element analysis and simulation
Stress, vibration and dynamics checked before metal is cut.
- Electrical · 2004
Electrical and control systems
Power, drives, wiring and controls for machines that run all night.
- New · 2027
Robotics integration
Mobility, arm, gripper and storage brought together as one machine.
- New · 2027
Artificial intelligence and computer vision
Bud recognition, maturity classification and bloom-time prediction.
Selected work from Vaftsy CAE projects. Client names and project details are withheld.
The team
Vinaay Patil
Head, AI & Robotics
M.S. Mechanical Engineering, University of Texas at Arlington. 18+ years in engineering and simulation. Author of First Steps in Artificial Intelligence and First Steps in Python.
- Mechanical design & CAE
- Ashutosh Dhumbre, Aniket Shinde
- Electrical & controls
- Ganpat R. Patil
- AI, vision & integration
- Vinaay Patil
- Field operations
- Shreyash Jagtap
- Founder, Vaftsy
- Rahoul Jagtap
Development journey
From concept to field trials, in the open.
We are at the start. Nothing has been built yet, and no robot is in commercial use. We will mark each stage complete here, with evidence, as we reach it.
- 1Ongoing · 2026
Concept and research
Studying the night harvest with growers; defining the machine and its subsystems.
- 2Planned
Data and AI
Night images of buds at every stage; first maturity and bloom-time models.
- 3Planned
Design and simulation
CAD of chassis, arm and gripper; structural and dynamic simulation.
- 4Target · month 9
Working prototype
Mobility, vision, manipulation and control integrated; bench trials.
- 5Target · month 15
Field trials
Pilot plots with growers: output, picking accuracy, damage and uptime measured.
- 6Target · month 24
Commercial system
First paying growers, once the economics are proven in the field.
Development updates
- Nov 2026Grower conversations near Pune on night harvesting, labour and losses. Robot concept and subsystem architecture defined.
- NextStart of the night-image dataset and first CAD of the straddle chassis.
How to read the timeline
Months count from the start of the funded programme in 2027. Stages 4 to 6 are targets, not commitments, and will move if the evidence says they should.
CAD screenshots and test photographs will appear here as each stage produces them.
Where we are going
One crop first. Then the whole field.
We are focused on one hard problem. The platform is being designed so that more tasks and crops could be added later as separate modules.
Nishigandha Harvesting Robot
A straddle-type machine designed to drive the bed at night, find buds that will open in transit, and pick them into on-board storage.
- Target~23 kg per 4-hour shift
- TargetBelow 4% damage
- Target₹2.5 lakh cost per robot
- StageConcept and research
Proposed capability: two to three robots per field could work under one central coordinator, which would assign rows, track progress and log each night's harvest. This is a design direction, not an existing system.
Possible later modules · not in development
Spraying and pest control
A spraying module for the same chassis, guided by the vision system.
Fertiliser application
Precise dosing along the bed, between harvest nights.
Other delicate crops
Other flowers grown in beds, once nishigandha is proven.
VAFTSYVAFTSY AI & ROBOTICSNishigandha Harvesting Robot
Work with us
Three ways to get involved.
Email us and mention which of these fits you best. We reply from the Vaftsy AI & Robotics team in Pune.
Growers
Explore a field pilot. Host a pilot plot and help shape how beds are laid out for a robot.
Suggested subjectField pilot enquiry
Research partners
Collaborate on technology validation, bud maturity and post-harvest behaviour.
Suggested subjectResearch collaboration
Investors
Discuss the development opportunity: prototype development and field trials.
Suggested subjectInvestment discussion
Emailai@vaftsy.com