Tanisha Gupta

About

Embodied AI/robotics-focused ML engineer who built a real-time assistive Braille device by coupling transformer OCR with an Arduino-controlled electromechanical Braille cell, solving tight latency and hardware-integration constraints. Has recent work on geometry-grounded world models and a real-time 4D reconstruction foundation model (Any4D), and delivered measurable impact at Voxel AI by building a scalable headless simulation + synthetic data pipeline that improved a safety-critical algorithm’s recall by ~16%.

Key Strengths

• End-to-end AI-to-hardware integration (OCR to electromechanical Braille cell control)
• Real-time/low-latency optimization for embodied systems
• Robust system design for uncertain ML outputs via deterministic post-processing
• Reliable embedded communication design (buffering + handshake protocols) to prevent dropped/misaligned outputs
• Designing efficient 4D representations for real-time reconstruction (Any4D)
• Advanced transformer/attention architecture design (alternating global and view-specific attention)
• Research-to-production decision-making using benchmarking (GUI vs headless simulation)
• Built scalable headless simulation + synthetic data pipeline
• Improved recall of a safety-critical algorithm by ~16% using synthetic data pipeline

Experience

Education

Skills

Languages

Publications

2 publications