Dhruv Chaudhary

How every project on this page ran

The same loop each time. State the target as a number before touching CAD. Simulate it, then check the simulation against a physical test. Design for the process that will actually make the part. Treat every missed target as the next design input rather than a failure to hide.

That last part is why each project below ends with a table that marks objectives as met or missed, in public. The misses are usually where the interesting engineering is.

Co-founder and lead engineer. Electricity-free refrigeration for the vaccine and medical cold chain.

The problem

Vaccines and temperature sensitive medicine have to stay between 2 and 8 °C. The stretch of the cold chain where that fails is the one running on unreliable mains power.

Both field workarounds fail in their own way. A compressor needs power that is not there. Ice packs sit below the window and freeze the payload instead of protecting it.

The hypothesis

An ionochaloric endothermic reaction, buffered by a phase change medium, can hold 2 to 8 °C with no compressor, no power and no consumable refrigerant.

I own product engineering, CAD, simulation, the bill of materials and IP strategy. My co-founder Mridul Jain owns chemistry, testing and the regeneration protocol.

How the system works

The payload sits inside a sealed cavity wrapped in removable cooling cartridges, a structural shell and a vacuum insulation layer. The cartridges are charged outside the unit in a reusable salt bath, installed, and then carry the payload passively for the length of the run. Nothing is plugged in and nothing is thrown away. The salts are recovered and returned to the bath, so water is the only consumable per cycle.

What the tests measured

Testing ran in stages: chemistry first, then the cooling medium at full volume, then the integrated unit. The final series closed in June 2026 with the integrated prototype holding the payload cavity inside the 2 to 8 °C window for more than 70 hours at 30 °C ambient, comfortably past the design duration.

Two products, one platform

A single envelope could not carry a three day hold and still stay light enough for a delivery rider, so the platform was split rather than compromised. Both tiers share the same reaction chemistry and the same reusable recharge model. Switch between them below.

Three design decisions that mattered

The architecture that survived is the third one. The first put the cooling medium in a fixed jacket around the payload. It hit the temperature target and lost on the workflow around it: recharging the whole unit took most of a working day, which is time no clinic or logistics operator will absorb. The medium moved out into sealed cartridges that are removed, charged outside the box and reinstalled.

The second was internal geometry. Recharge time scales with the square of the conduction path, so shortening that path was worth more than any change to the chemistry. Turnaround fell by roughly an order of magnitude with no loss of hold time, at the cost of some added mass.

The third was insulation. Vacuum panels beat foam by around ten times per unit thickness, but they cannot be trimmed without destroying the panel, so they are ordered at final size. That decision locked the outer dimensions of the product before anything else was fixed.

Results against target

Recognition

• The Earth Prize 2025, Asia Winner. Top 7 of 15,000 participants globally, with a $12,500 development grant.
• Conrad Challenge Innovator, top 15% worldwide.
• Indian patent application filed on the cooling method. DPIIT recognised startup.
• Keynote at the Nasscom Design and Engineering Summit 2025, and speaker at TEDxGateway.
• Expression of interest from a leading Indian vaccine manufacturer for pilot evaluation.
• Featured in over 50 international media outlets, including Business Insider and Fox News.
• Nominated for the Pradhan Mantri Rashtriya Bal Puraskar.

What I took from it

• The binding constraint was never the physics. Both earlier architectures cooled correctly and lost on the workflow around them, which is the part a simulation does not show you.
• The cost ceiling shaped the engineering as much as the thermodynamics did. The insulation decision is a price decision first.
• Learning what belongs in a patent application and what stays out of it. The disclosure boundary is an early design decision, not a late one.
• Research closes when the tests stop changing the design, not when the ideas run out.

Team principal and chief design engineer, Maverick Motorsports, STEM Racing India, formerly F1 in Schools.

I owned the aerodynamic design and manufacture of a CO₂ powered scale race car built to a locked regulation set, while running the team budget, sponsorship and delivery across engineering, marketing and operations to a fixed competition deadline.

Design targets

• Race time of 1.1 to 1.3 seconds over a 25 m track
• Mass at the 50 g regulation minimum
• Full compliance with the competition regulation set

Computer aided design

• Fusion 360 and SolidWorks, iterative modelling across seven full vehicle revisions
• Lofts, sweeps, shelling, assemblies and tolerance checks
• Zebra and curvature analysis to confirm manufacturability before committing to stock

Four prototypes, compared

Every revision was measured on the same five numbers. Pick one to see where it won and where it paid for it.

Invictus Tonitrum Nimbus Tempest

Tempest

Simulation and race time prediction

Drag force, flow uniformity and pressure drop were tracked across simulation iterations in Ansys Discovery and SimScale, alongside deformation, Von Mises stress and shear elastic strain. An Euler integration model converted drag coefficient into a predicted race time, which is what made the aerodynamic trade offs comparable to each other.

First innovation: tilted sidepods

The idea came from full scale Formula 1, specifically the shallow sidepod dip used to stabilise airflow. The tilt guides air outward and rearward, reducing pressure buildup, keeping flow attached and uniform, and improving the efficiency of the central tunnel and diffuser while easing the transition into the rear wing.

CFD showed a clear improvement in the uniformity of airflow coming off the front wheels once the tilt was applied. The gain was largest immediately behind the front wheel, which is where the wake is dirtiest and where the diffuser has the least tolerance for disturbed flow.

The constraint turned out to be manufacturing rather than aerodynamics. A steeper tilt performed better in simulation but produced an overhang that could not be machined cleanly in a single setup, and would have needed manual finishing on every unit. The geometry carried forward is the steepest angle that still came off the machine to tolerance. That is a trade made in favour of repeatability across seven builds rather than a single best lap.

Second innovation: rear extrusions

Small pyramidal extrusions sit near the tail of the car, surrounding the pressurised CO₂ canister. The team called them the teeth. The inspiration came from a land speed record car that used a similar surface treatment to manage its wake.

The teeth manage the negative pressure drag generated behind the car. Normally the turbulent wake left behind is uncontrolled and creates a low pressure zone that works against the car. The extrusions convert that turbulence into organised eddies and vortices which turn back into the body, recovering part of the loss as forward propulsion. A penalty was turned into a small gain.

Results against target

Recognition

• Nationals: Most Sustainable Team, Top 20 overall, Top 5 knockout time, fastest reaction race car time.
• Regional Champions, with the highest scores in Design and Engineering Portfolio, Verbal Presentation, Car Engineering, and fastest average race time at 1.13 seconds.
• Raised over $2,000 in sponsorship and managed the full team budget.

What I took from it

• Where simulation accuracy has to give way to manufacturability, and how to tell which side of that line a decision falls on.
• How to treat a missed target as a design input. The rear extrusions exist because the wake could not be removed.
• Closing the loop between aerodynamic theory and physical wind tunnel testing, rather than trusting either alone.
• Leading as team principal meant blending engineering judgement with sponsorship, budget and people.

Process automation intern, Rayfeild Systems, via the Harvard Undergraduate Ventures TECH Summer Program.

The brief was an AI assisted automation platform for renewable energy workflows, covering anomaly detection, forecasting and plain language reporting for the solar and wind sector.

Who it was built for

Before writing anything I mapped two personas with opposing needs, one technical and one operational. Every feature decision after that traced back to a documented pain point rather than an assumption. An energy analyst in Bangalore fighting scattered data across Excel, SCADA and reports. A maintenance supervisor in Texas drowning in irrelevant alerts with field teams who had no mobile access to live data.

What shipped

• Data cleaning and validation covering unit consistency, schema fixes and timestamp alignment
• Forecasting models, Prophet, XGBoost and regression, integrated with a weather API
• Anomaly detection using isolation forests over SCADA logs
• Plain language generated summaries for ESG and compliance reporting
• Zapier automation for alerts and workflow triggers

Outcome

• Selected in the top 15% globally for the Harvard Undergraduate Ventures TECH Summer Program
• Prototype validated against all three target personas: analyst, operations and compliance
• Direct feedback from the founder of Rayfeild Systems on the tool's potential for real deployment
• Completed 30 expert led sessions on venture capital, product strategy and market analysis

The clearest lesson: designing for a non technical persona is a harder constraint than designing for a technical one, and it changes the product rather than just the interface.

Three school research projects, each carried through to a tested prototype with a measurable outcome.

STUBBurn

Reduce stubble burning and the carbon emissions that follow from it

Approach

Convert crop residue into biomethane for clean fuel, and into bioplastics as an alternative to conventional plastic.

Result

Generated methane sufficient for lighting, and a bioplastic degrading in roughly 100 days.

Outcome

Best Overall Performance at Sahodaya Bal Vigyan.

Paripoorn

Prevent the carcinogenic reuse of waste cooking oil

Approach

Convert used oil into biodiesel and glycerine, then into downstream derivatives including contact lenses and fireproof bricks.

Result

Biodiesel with clean combustion, with glycerine and byproducts extracted successfully.

Outcome

Best Overall Performance at Sahodaya Bal Vigyan.

Vidyutottam

Prevent electrocution from leakage currents in electric poles

Approach

A low cost warning system combining an indicator LED with a fuse triggered mechanical danger sign.

Result

Working prototype. The LED illuminated and the danger sign deployed reliably under simulated leakage.

Outcome

State Level Finalist, CBSE Regional Science Exhibition.

Across all three: apply core chemistry, physics and engineering to problems that already exist rather than invented ones, and treat affordability and scalability as design requirements rather than afterthoughts.

Running programs, conferences and teams, and learning to explain technical work to people who did not ask for the technical version.