Mid-Level Research Scientist (Hybrid)

Serigor Inc. • Full-time • Boothwyn, Pennsylvania, United States • $0 / year • 6h ago

Job Title:

Mid-Level Research Scientist

Location:

Hybrid to Boothwyn, PA 7 Creek Parkway, Suite 710, Boothwyn, PA 19061 (3 days onsite) Duration:

6-month Contract-to-Hire. Employment Type on conversion: Full time; Exempt. Reports To: Chief Technology Officer

Job Description: We are seeking a skilled and motivated Mid-Level Research Scientist to join our team. The ideal candidate will focus on developing and deploying multimodal machine learning models specifically for speaker identification and verification tasks. This role involves designing and refining neural architectures that encompass various features, training and evaluating deep learning models, and enhancing the robustness of these systems for real-world applications in voice authentication and behavioral analysis. Job owner's addition: these individuals will be working with the production side of the house, working on existing models, day to day research, algorithm models, mathematic skills; they are heavier in PyTorch. Must be able to reach Top Secret Security Clearance the site is trying to get cleared for more government contracts as well. Standing calls 1 2 days a week.

Key Responsibilities: Model Development: Design innovative neural architectures that integrate speech, acoustic, and linguistic features for speaker identification and verification tasks. Data Handling: Train deep learning models on large-scale datasets, including participation in the construction and annotation of specialized datasets, such as the "American Dream Dataset". Evaluation & Benchmarking: Benchmark age prediction and speaker verification models, leveraging datasets to enhance model performance and demonstrate superior generalization. Research Prototyping: Conduct research initiatives focused on cross-modal representation learning and predictive modeling of political career advancement using voice quality and prosodic features. Optimization: Optimize existing models, including the development of lightweight architectures for resource-constrained environments, such as real-time image captioning systems. Architecture Design: Evaluate and benchmark diverse adapter architectures for vision-text alignment, while achieving state-of-the-art performance metrics on established datasets (e.g., COCO dataset). Collaboration: Collaborate with cross-functional teams to translate research findings into scalable solutions and real-world applications.

Minimum Education and Experience (client's Required list): Master's or PhD in Computer Science, Electrical Engineering, or a related field. (PhD preferred per the intake form.) 3-5 years of experience in machine learning and deep learning, with a proven track record of developing multimodal models. Strong proficiency in programming languages such as Python and frameworks including TensorFlow and PyTorch. Experience with acoustic and linguistic feature extraction and understanding of speaker identification and verification systems. Familiarity with natural language processing (NLP) and computer vision integrations, particularly in real-time applications. Strong analytical and problem-solving skills, with the ability to work independently and as part of a team. Excellent communication skills to present complex technical concepts to diverse audiences. US Citizen required for the client's government contracts and the site's clearance process. Must be able to reach Top Secret Security Clearance. No visa sponsorship.

Preferred Qualifications: Publications in relevant conferences or journals. Experience in research involving behavioral analysis and authentication systems. Understanding of model efficiency and optimization strategies for deploying machine learning models in production.

Soft Skills: Strong analytical and problem-solving skills. Able to work independently and as part of a team. Excellent communication skills able to present complex technical concepts to diverse audiences. [The intake form's soft skills field was left blank; the above is taken from the client's required qualifications.]

Working Conditions (as stated in the client JD): This position follows a hybrid work schedule based in Boothwyn, Pennsylvania. Employees are expected to work both remotely and onsite, with specific in-office days determined by departmental needs and supervisor approval. The onsite work environment is a professional office setting equipped with standard office technology and resources. Remote work requires a reliable internet connection, a private workspace conducive to productivity, and availability during standard business hours (Monday through Friday, 8:00 AM 5:00 PM EST). Occasional evening or weekend work may be required to meet project deadlines or support organizational priorities. The role may involve periodic travel for meetings, training, or collaborative projects. Reasonable accommodation will be provided to enable qualified individuals with disabilities to perform essential job functions.

Skills: Skill Required/Desired US Citizenship hard gate, due to government contracts and site clearance. Required

Able to reach Top Secret Security Clearance hard gate. Required

Master's or PhD in Computer Science, Electrical Engineering or a related field (PhD preferred) hard gate. Required

Machine learning and deep learning. Required 3-5 Years Developing multimodal models proven track record. Required

Years PyTorch the client is heavier in PyTorch than TensorFlow. Required

Years TensorFlow. Required

Years Python. Required

Years Acoustic and linguistic feature extraction. Required

Years Speaker identification and verification systems. Required

Years Designing neural architectures integrating speech, acoustic and linguistic features. Required

Years Training deep learning models on large-scale datasets; dataset construction and annotation. Required

Years Model evaluation and benchmarking speaker verification, age prediction. Required

Years Natural language processing (NLP). Required

Years Computer vision integrations, particularly real-time applications. Required

Years Algorithm modelling and strong mathematics. Required

Years Working on existing production models not greenfield research only. Required

Years Model optimization lightweight architectures for resource-constrained environments. Desired

Years Cross-modal representation learning. Desired

Years Publications in relevant conferences or journals. Desired

Research involving behavioral analysis and authentication systems. Desired

Years Model efficiency and optimization for production deployment. Desired

Years Voice biometrics / synthetic voice and deepfake detection background (e.g. from companies such as Pindrop or ElevenLabs). Desired

Years