AI Scientist/Engineer for Multimodal, Low-Latency Systems

SHAHAB JALALVAND, PHD

Summit, NJ • 973-***-**** • **********@*****.*** • Linkedin.com/in/Shahab-Jalalvand • shahabjld.com • GoogleScholar AI SCIENTIST/ENGINEER

LLM-Driven Innovations Data-Centric AI Speech-Language Intelligence Driven machine learning leader with extensive experience designing and deploying advanced generative AI, NLP, and ASR solutions for customer-facing environments. Adept at building zero-day models and refining data-intensive architectures to enable seamless, real-time transcription and redaction of sensitive information. Skilled at leading PhD-level research teams, accelerating speech-to-text accuracy, and integrating multimodal data for scalable enterprise applications. Known for producing high-performance, low-latency AI models that meet stringent industry standards. History of guiding strategic initiatives for cross-sector clients, from healthcare to finance, enhancing operational performance and shaping groundbreaking AI-driven offerings. Published extensively in top-tier conferences and recognized with multiple awards for research excellence.

• Multimodal AI Integration

• Large Language Model Fine-Tuning

• Neural Acoustic Modeling

• Domain Adaptation

• Zero-Day Model Deployment

• Low-Latency Inference

• Speech-to-Text Benchmarking

• RLHF Techniques

• Cloud-Based Model Orchestration

• Prompt Engineering Strategies

• Continuous Model Optimization

• AI Safety & Compliance

PROFESSIONAL EXPERIENCE

SJ AI SERVICES SUMMIT, NJ (2025)

2025: FOUNDER

Led the design, development, and deployment of production-grade AI solutions for clients across healthcare, e-commerce, and media. Worked cross-functionally with investors, business stakeholders, sales, marketing, IT, and healthcare providers to deliver end-to-end AI systems, from problem definition and research to deployment, monitoring, and commercialization. Projects: Ø SwayMind (Healthcare AI Platform): Led end-to-end development of a multimodal AI platform for mental-wellbeing awareness and triage. Designed and implemented voice-based conversational AI agents, speech and language analytics pipelines, and scalable cloud infrastructure. Collaborated with clinicians, healthcare providers, and compliance stakeholders to align AI outputs with clinical workflows and regulatory constraints. Worked with investors and advisors on product strategy, roadmap, and validation.

Ø EcommAgent (E-commerce Automation Platform): Architected and delivered a multimodal AI system for automating online product listings, dynamic pricing, and promotional content generation. Integrated LLMs, computer vision, and market intelligence signals to support automated video creation and weekly price optimization. Partnered with sales and marketing teams to align AI capabilities with revenue growth and customer acquisition strategies. Ø Transcribed (Media & Entertainment AI Pipeline): Designed an end-to-end transcription and post-production analytics service for reality TV and long-form video content. Implemented high-accuracy ASR pipelines with speaker diarization, fine-tuned LLMs for highlight extraction, and computer vision models for automated burned-in timecode analysis. Worked closely with production teams and IT stakeholders to integrate AI outputs into existing editorial workflows. PING DATA TECHNOLOGY INC SUMMIT, NJ (2025)

2025: AI CONSULTANT

Led advanced AI initiatives focusing on insurance document processing, specializing in Excel sheets and PDFs. Ø Designed and deployed a Deep Neural Network (DNN) solution for processing large-scale tabular data, optimizing efficiency and accuracy.

Ø Implemented and managed sophisticated Git workflows, ensuring seamless collaboration and version control across teams. Ø Leveraged AWS services, including Lambda functions, to develop scalable and resilient AI applications. Ø Contributed to strategic discussions and technical decision-making, enhancing project outcomes and client satisfaction. INTERACTIONS LLC NEW PROVIDENCE, NJ (2017 – 2024) 2021 – 2024: PRINCIPAL INVENTIVE SCIENTIST

Led a team of five PhD researchers in developing and deploying cutting-edge AI systems for real-time applications. Directed the enhancement of language models, supervised weekly sprint planning, and ensured delivery of impactful results to management. Served as a language model expert on high-profile projects, including Trustera, which focused on live-sensitive entity redaction. Ø Improved Trustera’s sensitive entity redaction AI system by 15% annually over two years, enabling $4M and $5M contracts in 2024 and 2025, respectively.

Ø Authored and coordinated the publication of a Trustera scientific paper, elevating project visibility in top-tier AI conferences and enhancing marketability.

Ø Engineered production-grade AI models for sensitive data redaction, processing 10M+ calls monthly with real-time accuracy. SHAHAB JALALVAND, PHD Page 2

Ø Benchmarked conversational entity tagging using LLMs, evaluating Google Gemini, OpenAI ChatGPT, Meta LLaMA, and Mistral for optimized AI deployments.

Ø Designed a reinforcement learning pipeline with SQL, Bash, and Python, enabling 50% faster model updates and real-time system monitoring.

Ø Developed low-latency, real-time ASR systems to improve call transcription and confirm compliance with privacy standards. 2017 – 2021: SENIOR INVENTIVE SCIENTIST

Analyzed and processed extensive call-center datasets to develop and train AI models to improve word accuracy and intent classification. Spearheaded the design and deployment of innovative AI solutions for call center automation, enabling optimized speech-to-text accuracy and real-time intent recognition. Ø Developed a zero-day modeling plugin to address new client requirements, reducing dependency on human analysts by 50% and saving $200K annually.

Ø Introduced advanced speech recognition technology, achieving a 20% improvement in word accuracy and reducing operational costs by $2M annually.

Ø Designed Transformer-based ASR systems tailored for Intelligent Voice Assistant (IVA) applications. Ø Processed billions of call-center records in SQL, generating domain-specific embedding models to boost system precision. Ø Benchmarked speech-to-text technologies across platforms, including AWS, Google Cloud, IBM Watson, Nvidia Nemo, and Nvidia Riva, identifying optimal solutions for deployment. Ø Fine-tuned large transformer models using Nvidia A100 GPUs, increasing computational efficiency by 50%. Ø Guided capstone projects at Rutgers University, resulting in a research award for pioneering work on live video classification in technical support scenarios.

COGNERA AI SUMMIT, NJ

2023 – PRESENT: DIRECTOR OF AI

Drive the strategic implementation of AI technologies to address complex business challenges. Collaborate with industry stakeholders to identify AI opportunities and lead innovative projects from conception to deployment. Represent the organization at technology forums and establish relationships with entrepreneurs and venture capitalists to secure partnerships and funding for AI initiatives. Ø Developed HaomaX, a healthcare platform capable of reducing radiology report generation time from 24 hours to five minutes by integrating multi-modal AI models, ResNet101, and LLaVA2 through prompt engineering. Ø Built and evaluated an advanced radiology reporting system with human-in-the-loop capabilities, ensuring compliance with AI safety metrics and accuracy standards.

Ø Validated AI-generated radiology reports with teleradiologists, fostering trust and driving adoption. Ø Introduced AI consulting services to a Maryland-based company transitioning to AI, guiding federal contract proposals that expanded client opportunities by 40%.

Ø Spearheaded development of live medical report dictation using NVIDIA Nemo’s fast-conformer model. Ø Created a custom WebRTC-powered video conferencing platform to support multimodal AI tasks, facilitating seamless collaboration across remote teams.

Ø Conceptualized and developed ConverseX, a mobile application for speech translation in 55 languages using Python Kivy, Xcode, and Google APIs.

ADDITIONAL EXPERIENCE

2016 – 2017: PERVOICE • RESEARCH ASSISTANT: Utilized advanced automatic speech recognition (ASR) techniques to enhance existing systems, including ASR quality estimation based on doctoral research. Conducted optimization of decoder systems to reduce computational requirements while maintaining high accuracy. Focused on developing ASR solutions for low-resource languages through web crawling and data preparation. Designed and evaluated models to increase system accuracy and efficiency, delivering optimized speech-to-text solutions for broadcast, call center, and reporting industries. 2012 – 2016: HLT-GROUP OF FBK • PH.D. RESEARCHER: Designed and executed experiments to improve transcription accuracy using innovative continuous-space language models and optimization techniques. Conducted advanced research in machine translation, speech recognition, and natural language processing under European projects, such as EU-Bridge and Mate-CAT. EDUCATION & COMPETENCIES

Licensed Inventor in Real-time AI Systems Columbia University, USA Focus: Real-time generation of radiology reports using multimodal generative AI Project Mentor in AI & Machine Learning Rutgers University Focus: Industrial applications of speech and language processing systems Ph.D. in Information and Communication Technology University of Trento, Italy Focus: Automatic Speech Recognition Quality Estimation SHAHAB JALALVAND, PHD Page 3

M.Sc. in Artificial Intelligence and Robotics Iran University of Science and Technology Focus: Speech Recognition and Acoustic Modeling

B.Sc. in Software Engineering Iran University of Science and Technology Focus: Design and Development of a Heart Rate Monitor Using Microprocessors Languages: English: Fluent Italian: Intermediate Persian: Native Arabic: Beginner Appointments & Awards: Best Paper Award, ICEE Conference, Iran – Recognized for the first publication during M.Sc. studies. Best Research Award, 2019 – Achieved for mentoring Low Resource Multilingual ASR Data project. SELECTED PUBLICATIONS

Ø E. Gouvea, S. Jalalvand, et al., "TRUSTERA: A Live Conversation Redaction System," IEEE International Conference on Acoustics, Speech and Signal Processing, IEEE, 2023. Ø K. Singla, S. Jalalvand, et al., "Seq-2-Seq Based Refinement of ASR Output for Spoken Name Capture," 23rd Annual Conference of the International Speech Communication Association, ISCA, 2022. Ø S. Jalalvand, et al., "Transcrater: a tool for automatic speech recognition quality estimation", ACL, 2016. TECHNICAL SKILLS

Ø GenAI Tools: OpenAI GPTs, Google Gemini, Anthropic Claude, N8N, LangChain, LangGraph Ø Programming Languages: Python, C++, Bash/Shell scripting, Visual Studio, Google Colab Ø Machine Learning Frameworks & Libraries: TensorFlow, PyTorch, Keras, Scikit-learn, Hugging Face Transformers Ø ASR & NLP Tools: Kaldi, CMU Sphinx, OpenAI APIs, SpaCy, NLTK, SRILM, KenLM, ASR Benchmarking Tools (e.g., TranscRater) Ø Data Processing & Visualization: Pandas, NumPy, Matplotlib, Seaborn, Jupyter Notebook Ø Cloud Platforms & Services: AWS, Microsoft Azure, Google Cloud Ø Development & Collaboration Tools: Git, Docker, Kubernetes, CI/CD Tools (e.g., Jenkins, GitHub Actions) Ø Research & Publishing Tools: LaTeX, Overleaf, Conference Management Systems (e.g., EasyChair, CMT) Ø General Software: Slack, Microsoft Teams, Zoom, Confluence, Jira, MATLAB

Contact this candidate