Kaustubh Krishna
ReportI'm Kaustubh, 20, based in Bengaluru and building my second company. I created Mekai Open and am its only core maintainer. I started coding at nine. Around ten, i joined my sister's Java classes, taught 3-4 seniors while learning and earned from it. At eleven, an electrician taught me to read a printed circuit and solder components onto a PCB. I went on to build my own PCBs, a 3D printer and robots. For World Robot Olympiad, i travelled three hours a day to the science centre to build. We became the first and only team from Bihar to qualify for nationals, in 2019 and again in 2022. At 19, i built a racing controller solo: 3rd globally in ICRA 2025 RoboRacer simulation qualification, ahead of teams from UPenn, UCF and Minnesota. Both my open-source Reinforcement Learning Toolkit and Industrial Control Toolkit were used in that work. I also built and deployed an autonomous golf cart on campus. In my first company, i built an engine for robot incident reconstruction, root-cause analysis and safety pre-compliance. Technical discussions with SICK led to free hardware and consultancy. I presented at ROSCon India and took a technical session for eInfochips' Head of Robotics and 30+ senior engineers. Now i build the platform at Mekai Labs and work directly with factory teams. Bhavyaa and i reached 6+ Indian OEMs within one month and have visited those OEMs and 40+ other manufacturing units. That gives us equipment and operating problems to build around. I want the reusable engineering from this work to reach people who don't have the same factory access. My earlier projects: https://github.com/kaustubhkr/reinforcement-learning-toolkit and https://github.com/kaustubhkr/industrial-control-toolkit
Mekai Open
Mekai Open is open-source infrastructure for connecting industrial data and building reusable tools around it. Its current public core reads recorded VDA 5050 robot messages, checks them and calculates operational results with the source and missing data attached. It has 13 registered functions. The VDA component is already used at one AMR-company client site. A lot of industrial devices are closed source. You can't properly build an adapter just from a public dataset. You need to get to the factory, connect to the actual equipment, understand its interface and test what happens. Most independent robotics builders don't have that access. We do, through our work at Mekai Labs. I want to turn more of that work into open adapters, interfaces and tools that other teams can install and extend. Start with the VDA core, make it usable outside our setup, then add more robot and controller interfaces as we test them. The wider roadmap includes reconstructing machine runs, comparing changes and checking how robots interact with the equipment around them. The rest of the Mekai Labs platform is private for now. Mekai Open should be useful on its own. The equipment can stay closed source while the adapter we build around its available interface is open. A robotics team should be able to use our work without repeating the same integration from scratch. The current public release is qualified with synthetic and lab tests. Use of the VDA component at one client site does not establish compatibility across vendors. The next work is to make the public interfaces, examples and tests usable by another team.
Mekai Open adapters and public release
INR 3 lakh for 3 months: 1.8 lakh for development and maintenance, 1.2 lakh for factory travel. Month 1: release the VDA analysis command, setup guide, examples and missing/out-of-order message tests. Month 2: build and test one more industrial-data adapter using factory equipment. Month 3: independent-team trial, fixes, contribution guide and public report. I'll share quarterly updates. No outside support received. The grant funds open components; private data and Platform stay separate.