class Prithweeraj:
    role      = "AI Researcher, Founder, Computational Artist"
    company   = "Co-Founder and CEO @ HaorGrix"
    school    = "CS and Engineering @ University of Toledo (grad 2028)"
    research  = ["AI Fairness", "Generative AI", "NLP", "Model Evaluation"]
    building  = ["Reiva", "Irris", "RoboTheater"]
    belief    = "I am a painter, and it is high time to rethink AI art"

• 📄 Published at IEEE Xplore (QPAIN 2025) -- multi-layer LSTM ensemble for Bengali fake-news detection | 1 Citation
• 🔬 Researching AI fairness and generative-model bias -- how the curation step shapes what text-to-image models learn
• 🚀 Co-Founder and CEO of HaorGrix -- an AI-native software company shipping real products from Bangladesh to the world
• 🤖 Built Reiva (AI receptionist for clinics) and Irris (a mobile AI agent that acts, not just answers)
• 🎭 RoboTheater -- placed a frontier vision-language model inside a live trolley-problem simulation to test perception against ethics
• 🎨 Generative, interactive art and the book Material: Painting in the Post-AI Era

Recursive Aesthetic Reinforcement (RAR) Ongoing research on aesthetic bias in generative AI An end-to-end audit of Stable Diffusion v1.5 against the LAION-Aesthetics V2 curation step. Base-rate-controlled, pre-registered, and reproducible on free compute.

Bengali Fake News Detection: A Multi-Layered LSTM Ensemble Approach QPAIN 2025 (IEEE Xplore) | Co-Author | 1 Citation | DOI: 10.1109/QPAIN66474.2025.11171927

Context-Aware Dialectal Arabic Machine Translation arXiv:2604.06456 | 2nd Author | Rule-based augmentation to 57k parallel sentences, fine-tuned mT5 for controllable region and register

DP-4DGS: Dynamic Priors for 4D Gaussian Splatting Ongoing research with Dr. Aniruddha Ray, University of Toledo | Dynamic scene reconstruction with physics-informed priors

Product @ HaorGrix | LLM Voice Scheduling Real-time

Answers every call, books the appointment, and never misses a lead. Patient chat, digital prescriptions, WhatsApp booking, and a live analytics dashboard in Bangla and English.

Your phone, on autopilot | Agents Planning Mobile Tool-use

A natural-language agent for the phone: give it a goal and it plans the steps and carries them out across your apps and device.

VLM in a Live Simulation | VLM Gemini Three.js Evaluation

Placed Google Gemini inside a real-time Three.js trolley-problem film. A Python loop queries the model with a zero-context prompt and renders its decision live. It reasons about ethics correctly but fails at perception.

Painting in the Post-AI Era | Generative Art WebGL Writing

A gallery-grade catalogue and a full-screen reader for the book Material, alongside the paintings it was written beside.

One-verb game | WebGL GLSL Interactive

A hold-to-hide one-verb game, all art and sound generated in code.

Generative, endless, interactive | WebGL GLSL Audio

A contemporary generative piece built from the whole body of work, musical and never the same twice.

📄 Published Researcher -- IEEE Xplore (QPAIN 2025) with 1 Citation

🚀 Co-Founder and CEO -- HaorGrix, AI-native software company

:frame_photo: Exhibited Artist -- Mind & Soul Gallery & Studios, Toledo (Group Exhibition, May 2026)

🏆 International Recognition for Visual Art -- Teravarna, 2025

🏸 Champion -- UTBSA Men's Badminton Tournament, 2024 and 2025

🎨 Mixed-Media and Computational Artist -- material-expressions.vercel.app

> System.out.println("Open to research collaborations, AI product work, and creative engineering.");