Summary
Tharak Nagaveti is a founder and 3D engineer building at the intersection of generative AI and real-time graphics. As founder and chief executive of Hydrilla AI, he leads Bluefox3D, a 3D reconstruction engine that turns text and images into production-ready assets for game studios and independent developers.
His work pairs applied research with shipped product: peer-review machine-learning publications on one side, a platform serving 1,000+ users and 10,000+ generated models on the other. Earlier projects span legal analytics and mobile product development, a consistent record of moving research-grade ideas into tools people use.
Experience
Founder & CEO
Hydrilla AIDeveloping Bluefox3D, a 3D reconstruction engine built on Rectified Flow Transformers and TRELLIS-2 architectures.
- Scaled the platform to 1,000+ users and 10,000+ generated models, and onboarded 5 professional studios.
- Architected production-ready text-to-3D and image-to-3D pipelines using 3D Gaussian Splatting (3DGS).
- Lead product, AI engineering, and the end-to-end 3D pipeline for game studios and independent developers.
Software Developer
Idea ParticlesBuilt a mental-fitness platform and owned its mobile product end to end.
- Took a Mental Fitness platform from zero to 1,000+ downloads within 30 days.
- Optimized cross-platform performance in Flutter, managing UI/UX and state architecture end to end.
Research & publications
Predictive Maintenance using Machine Learning
Engineered multi-class classification models (LSTM, gradient boosting) for real-time anomaly detection across IoT sensor arrays, cutting unplanned industrial downtime.
Predictive Modeling of Institutional Analytics via AI/ML
Applied ensemble methods and dimensionality reduction to surface the core performance determinants in institutional benchmarking (NIRF).
Projects
Bluefox3D
A 3D reconstruction engine built on Rectified Flow Transformers and TRELLIS-2, generating production-ready geometry from text and images with 3D Gaussian Splatting.
Deepfake Detection System
Combined spatial landmark tracking with DCT frequency-domain analysis to flag the artifacts left behind by high-fidelity synthetic media.
Bail Reckoner
An NLP-driven legal analytics engine that pairs case modeling with explainable AI (XAI) outputs, making its reasoning legible to non-technical users.
Hawan Research Labs
Research and engineering on 3D datasets, supporting the text-to-3D generation and reconstruction work behind Hydrilla AI.
Honors & recognition
Founders Inc Canopy Program (500 Startups)
Selected for the global cohort, a competitive program for early-stage founders.
Top 10 Startup, NIT Jaipur
Recognized among the top ventures from a field of 400+ competing startups.
Smart India Hackathon 2024
Finalist through internal selection.
Technical expertise
- 3D engineeringReconstruction pipelines, 3D Gaussian Splatting, topology, and production-ready geometry.
- Generative AIRectified Flow Transformers, TRELLIS-2, and diffusion-based text-to-3D and image-to-3D.
- Machine learningLSTM, gradient boosting, ensemble methods, and dimensionality reduction for applied research.
- AI product developmentTurning generative models into shippable, reliable products.
- Mobile & cross-platformFlutter product development with end-to-end UI/UX and state architecture.
Education
Undergraduate, Technology
IndiaPursuing an undergraduate degree in technology while building Hydrilla AI alongside his coursework.