I am currently working as a Research Engineer at BlackRock. My work here spans agentic multi-agent systems, and I have recently been working on risk-aware portfolio construction. I have previously worked on deep learning research at Intel, while being actively involved in building end-to-end AI-native applications throughout my time at university.
I am deeply interested in computer vision and how intelligent systems can solve complex real-world problems. On the side, I love to talk about anything related to finance, cricket, and F1 (my bet is Lewis wins this year's Grand Prix). If you'd like to discuss ideas, projects, or just want to take me up on that bet, feel free to email me or reach out on socials. I'm always open to thoughtful conversations and meaningful collaboration.
Current Thoughts
Photo and video galleries haven't evolved in years. The way we interact with our digital memories feels stagnant, and I believe we are overdue for a major overhaul. A modern gallery experience should be deeply collaborative, unlock complex actions with a single click, and offer hyper-personalization that anticipates our needs. As visual intelligence advances, these capabilities will become the foundation for the next generation of spatial computing and VR.
This concept is currently in the idea phase. I am actively looking for a passionate designer or visionary who can help imagine and visualize this future. If that sounds like you, I'd love to talk!
Selected Achievements
-
2025
BlackRock: Co-authored a U.S. Patent related to agentic risk-aware portfolio construction and intelligent order sizing.
-
2023
M# National Hackathon: Top 6 among 360 registered ideas for "GROOT" - an app designed to encourage sustainable transport with functional AR gamification.
Publications
Riemannian Geometry: Speech Detection from MEG Brain Signals towards Non-Invasive BCI
Priyanshu Vij, Aryan Mangla, Raj Mehta, Jonas De Schouwer
Experience
Research Engineer
Designed and implemented an LLM-powered multi-agent system for equity portfolio construction. Engineered and optimized data & ML pipelines together with evaluation framework, reducing end-to-end latency and implementing strict guardrails.
Research Intern
Designed and implemented a one-stop agentic workflow in Aladdin for index subscription, reducing average index subscription time by 80%.
ML Research Program
Developed a CNN model for image classification using MNIST dataset. Deployed quantized INT8 model with 2.1x faster inference than FP32 while maintaining 1% accuracy drop. Leveraged Intel's TensorFlow optimizations to outperform benchmark metrics.
Projects & Explorations
Built an AI-native, self-iterative video refinement system that generates Hollywood-caliber cinematic videos from user prompts. Implemented an auto-iterative loop with background jobs and SSE real-time progress for long-running video generation.
Designed a tool to convert GST invoices from PDFs & images into structured, searchable formats. Built a chunking and stepwise text-extraction pipeline using early GPT-style models for reliable data parsing. Deployed for a pilot test with over 30 Indian SMB vendors for faster inventory workflow.
Developed custom diffusion pipeline supporting text+sketch conditioned generation with LoRA adapter support. Enabled training on consumer hardware by implementing Gradient Checkpointing and FlashAttention-2, reducing peak VRAM requirements by 35%.
Built and launched a cross-platform student information/events app and admin web portal used by over 10,000 student users. Designed completely customized access-based dashboard interfaces for moderation.
Co-Founder
An early exploration of how future AI systems could interact with spirituality and devotional practices. Built a voice-led devotional idol prototype, imagining how people might engage with spiritual guidance and rituals in a post-AGI world.
Designed AR gameplay where rare "Groot" spawns appear only on green routes & shared-transport nodes. Selected Top 6 / 360 at a national hackathon for functional AR gamification.
Technical Skills
- Languages: Python, C++, C, JavaScript, HTML, CSS, CUDA
- Frameworks: React, React Native, Next.js, Node.js, Express, Flutter
- ML Frameworks: PyTorch, JAX, TensorFlow, Keras, HuggingFace
- Cloud & Tools: Git, Linux, Docker, AWS, SQL, MongoDB, Prisma, Supabase
Education
CGPA: 8.00