Eric Soderquist Email & Phone Number

Champaign, Il, Us

Champaign, Illinois, United States

• Launched a free AI-powered voice translation platform supporting over 50 languages, facilitating real-time cross-linguistic communication for more than 1,000 users, with a focus on empowering immigrants, travelers, and ESL learners.• Designed and deployed custom multimodal language models, utilizing Reinforcement Learning from Human Feedback (RLHF) and divergence training to deliver adaptive, context-aware translations.• Engineered and maintained a cloud-native infrastructure leveraging Azure, TypeScript, React.js, Next.js, and PostgreSQL, ensuring high scalability, reliability, and availability. Secured cloud funding through a strategic Microsoft partnership to support platform growth and global reach.

Champaign, Illinois, United States

• Designed an AWS Permissions Automation system, reducing an 8-hour manual task to a 1-second process, enhancing productivity and security compliance.• Built a scalable distributed web-crawling system with Docker and Kubernetes, reducing data acquisition time from months to 2.5 hours and ensuring continuous updates, optimizing dealer network strategies for U.S. and UK competitors.• Developed an NLP microservice on Amazon EC2 Linux to parse complex addresses, resolving a critical blocker, saving costs by avoiding third-party solutions, and improving data pipeline reliability.• Automated warranty ID comparisons with a GUI-based tool, reducing a two-week manual process to 90 seconds and eliminating errors.• Deployed and optimized ETL pipelines across AWS, ensuring scalable data transformation and visualization through S3 and VistaMap, with high availability and fault tolerance.

San Francisco Bay Area

• Optimized Gemini Ultra's performance, boosting MMLU benchmark accuracy from 78% to 90.04% and HumanEval benchmark accuracy from 37.6% to 74.4%, significantly enhancing its multilingual capabilities and proficiency in coding tasks.• Applied divergence-based RLHF to improve model alignment for coding and language applications, ensuring highly adaptive and contextually relevant responses in real-world scenarios.

San Francisco Bay Area

• Designed and optimized large-scale data preprocessing pipelines, increasing data processing efficiency by 35% during the transition from GPT-3 to GPT-4, enabling smoother deployment and scaling.• Enhanced data validation and sanitization protocols, reducing hallucinations by 82% and improving model reliability and factual accuracy by 40%, ensuring higher-quality outputs in production.• Strengthened model safety and alignment, achieving an 82% reduction in disallowed content generation with GPT-4 compared to GPT-3.5, enhancing compliance and user trust.

Champaign, Illinois, United States

• Developed deep learning architectures for multi-modal neuroimaging fusion (MRI/EEG/fNIRS), enhancing neural decoding accuracy by 45% and improving responsiveness in real-time brain-computer interface (BCI) control systems. Designed parallelized preprocessing pipelines, reducing computational overhead by 60%, enabling faster data processing and analysis for high-dimensional neurophysiological datasets.

Deer Park, Illinois, United States

Assisted Living & Memory Care