Reads what your team actually produces
STEP, IGES, STL and DXF geometry, FEA and CFD output, PDF reports, Markdown notes and whole directory trees.
.step.iges.stl.dxf.pdf.md.csv
Ask your engineering history anything.
Parikalp is AI search for hardware engineering teams. It reads your CAD files, simulation reports and design reviews, so anyone on the team can ask why a part is the way it is and get the answer with its sources.
Now onboarding engineering teams for the closed beta. Works alongside Fusion 360 today, with SolidWorks next.
Knowledge gets trapped inside the tools that made it.
Every engineering decision leaves a trace. That trace is buried in a proprietary file format, on a drive three projects back, owned by an engineer who left in 2021.
-
Scattered files
Critical files live inside CAD tools, disconnected drives and departmental folders with no common index. Finding anything means knowing where to look, and most engineers don't.
-
Reasons that walk out the door
When an engineer leaves, the reasoning behind a tolerance, a material choice or a failed test leaves with them.
-
Hours lost to searching
Senior engineers spend hours combing directories. Every one of those hours is taken from the problem they were hired to solve.
What searching costs your team
$2,592,000
a year spent finding work that already exists, over 48 working weeks. That's 7.5 full-time engineers.
One layer on top of the tools you already use.
No migration and nothing to rip out. Parikalp reads from your existing drives, CAD and PLM systems and makes all of it searchable.
- 1
Ingest
Connect drives, CAD vaults and PLM exports. Nothing moves.
- 2
Parse
Geometry, metadata, simulation output and review notes are read file by file.
- 3
Connect
Parts, materials, tests and decisions are linked into one knowledge graph.
- 4
Ask
Anyone can ask in plain English and get an answer with its sources.
Understands how things relate
Goes past keyword matching. Parikalp knows that a bracket belongs to a chassis, that a test report validated a material change, and who signed off on it.
Answers with receipts
Ask "why did we switch the chassis alloy in Q3 2022?" and get the simulation report, the design review notes and the engineer's reasoning, each linked to its file.
Try it on a sample archive.
This demo searches a sample archive of 27 files from a made-up hardware company: design reviews, FEA and CFD reports, failure reports, drawings, test logs and cost sheets. Pick a question or ask your own in plain English.
Answers show up here, with every file they came from.
Sources
Answer
You're looking at sample data. Request early access to ask these questions of your own files.
See all 27 files in the sample archive
Built first for teams where precision is non-negotiable.
We're starting with engineering teams of 20 to 500 people in three industries where designs are complex and lost knowledge is expensive.
Aerospace and defense
Simulation-heavy validation and strict certification paperwork create the deepest silos. Every tolerance has a story, and Parikalp makes it findable.
Used for: finding FEA and CFD results, certification traceability
Parikalp for aerospace and defenseAutomotive
Platform programs run for years and pass through hundreds of engineers. Parikalp turns that history into something you can ask, not something you dig through.
Used for: platform reuse, supplier change impact
Parikalp for automotiveIndustrial machinery
Bespoke machines built over decades. When a component needs an upgrade, Parikalp surfaces the original design intent instead of a guess.
Used for: retrofit engineering, legacy documentation
Parikalp for industrial machineryNew hires get up to speed in days, not months. On their first day they can ask the whole company's design history the questions they'd otherwise save up for a senior engineer.
Your designs never leave your control.
Design data is the most sensitive thing a hardware company owns, so you choose exactly where it lives.
- Run it where you want
- Teams can use our cloud-hosted models. Enterprise customers can run Parikalp entirely on local models inside their own infrastructure.
- Never used to train shared models
- Your CAD files, simulation results and design notes stay in your deployment. They're never shown to other customers or used for training.
- Encrypted in transit and at rest
- Every file and every question is encrypted. On on-premise deployments, nothing crosses your network boundary at all.
- Access by role, with an audit trail
- Decide who can ask about which projects, and see a log of every search, so sensitive programs stay with the people cleared for them.
Built by engineers who felt the problem firsthand.
We spent years around hardware teams watching hard-won knowledge disappear. Parikalp is the tool we wished existed.
-
Aarsh Jain
Founder and CEO
Sets the product and architecture, bringing mechanical first principles to how AI is deployed. Loves robotics, aerospace and mechatronics.
Mechanical engineering, advanced CAD, IIT Guwahati
LinkedIn -
Durgesh Shelke
Tech lead
Connects software engineering with complex CAD data structures and keeps the product infrastructure automated and solid. Former Google intern and e-Yantra Robotics finalist.
Computer science, systems automation, IIT Guwahati
LinkedIn -
Raghav
Tech lead
Leads the development, training and deployment of the ML models, and designs the pipelines that process heavy engineering datasets.
Mechanical engineering, AI and ML, IIT Guwahati
Advisors
-
Prof. Sajan Kapil
Strategic advisor
Guides the mathematics and research behind features for complex engineering workflows.
Deep-tech R&D, CAD parsing, IIT Guwahati
LinkedIn -
Dr. Amit Kumar
Strategic advisor
Mentors the founding team on deployment, growth and building a company for a global market.
Venture scaling, AI products, Canada
Questions engineering leads ask us.
What is Parikalp?
Parikalp is an AI knowledge layer for hardware engineering teams. It connects to the drives, CAD vaults and PLM exports your team already uses, reads the geometry, metadata, simulation output and review notes inside them, and links parts, materials, tests and decisions into one knowledge graph. Anyone on the team can then ask a question in plain English, such as why a bracket changed material, and get an answer that cites the exact files it came from. It works alongside Fusion 360 today, with SolidWorks next, and runs either in our cloud or on local models inside your own infrastructure. Parikalp is built for engineering teams of 20 to 500 people in aerospace and defense, automotive and industrial machinery, and is in closed beta. The team also publishes research on finding existing CAD parts by their 3D shape.
Which CAD and engineering formats does it work with?
STEP, IGES, STL and DXF geometry, FEA and CFD output, PDF reports, Markdown notes and CSV data. Parikalp sits across your existing CAD and design tools rather than being a CAD kernel of its own: Fusion 360 works today, SolidWorks is next, and ingestion is built to extend to whatever formats your team standardizes on.
Does it replace our CAD or PLM tools?
No. Parikalp doesn't touch modeling. It runs alongside Fusion 360 today (SolidWorks is next) and your PLM system, indexing files, simulation results and design-change history so any of it can be found and asked about in plain English.
Can it run entirely on our own infrastructure?
Yes. Team plans run on our cloud-hosted models. Enterprise customers can deploy Parikalp on local models inside their own infrastructure for complete data isolation.
How long does it take to ingest our archive?
It depends on the size and structure of your archive. We scope it with you during onboarding and start with the projects your team searches most, so you don't wait on a full migration.
What does it cost?
Pricing depends on team size and whether you run in our cloud or on your own servers. Request access below and we'll put together a plan for your team.
Join the first cohort.
We're onboarding a small group of engineering teams for the closed beta. If your team spends real hours looking for past work, we'd like to talk.
What happens next
Or talk to the founder directly
Every message goes straight to Aarsh Jain, not to a queue.
Request received.
We'll email you within 48 hours to set up a call. Keep an eye on your work inbox.