Ishika Kulkarni

IBM · Lucknow, India

• Built an image preprocessing pipeline that enhances scan quality for OCR engines such as PaddleOCR, targeting noisy, low-contrast medical prescription and clinical document images.
• Applied computer vision techniques including denoising, adaptive contrast enhancement, deskewing, and binarization to improve text region clarity before recognition, contributing to a more robust end-to-end medical OCR pipeline.

CORES Lab, UCLA · Advised by Prof. Danijela Cabric (IEEE Fellow)

• Developing ML-assisted compressive phased-array radar algorithms for multi-target beam tracking in mmWave 5G/6G wireless systems.
• Designed attention-based neural networks and implemented compressive beamforming and sparse sensing pipelines that reduce beam-alignment overhead versus exhaustive scanning methods.
• Contributing to FPGA radar processing pipelines for mmWave sensing and beam-tracking validation.

Embedded Machines & Ubiquitous Robotics Lab, UCLA · Advised by Profs. Ankur Mehta & Anushri Dixit

• Engineered an amphibious quadrupedal robot with integrated sensing and control systems for multi-terrain navigation.
• Integrated embedded sensors and electromechanical components, significantly improving environmental perception and system robustness.
• Developed CAD-based designs and iterated prototypes to optimize performance in land and water environments.

Visual Machines Group, UCLA · Advised by Prof. Achuta Kadambi

• Evaluated LLM- and VLM-based coding agents on computer vision tasks, benchmarking task success rate, execution steps, and token efficiency.
• Researched and implemented Interaction-Aware Layered Generation using layered diffusion that enforces cross-layer consistency and improves editability by nearly 30%.

Atob Fuel Card · San Francisco, CA

• Collaborated with cross-functional teams to build a real-time fraud-detection ML model on transactional and behavioral data; queried and processed large-scale datasets in SQL and Python to support model development and deployment.
• Built the company's first end-to-end risk-scoring pipeline to predict repayment likelihood and enable proactive intervention for overdue payments, reducing false positives while improving risk classification accuracy.

Project Digital Design, Architecture & Verification · IEEE at UCLA

• Designed a modular FPGA game system in SystemVerilog, integrating synchronized display output and diverse I/O peripherals.
• Engineered timing-critical logic across multiple clock domains, developed comprehensive testbenches, and optimized hardware resource usage within a collaborative team.

UCLA Residential Life

• Lead event programming, peer mentorship, crisis response, and wellness and inclusion initiatives for a residential community.

UCLA Physics 5BL

• Facilitated hands-on lab sessions on thermal physics, fluids, waves, optics, and biochemical systems for a section of 30 students.