Data Scientist

Hybrid Onsite in Houston, TX : 2 days a week onsite, 3 days WFH

Relocation Assistance: Offered

Must be authorized to work in the US, will not sponsor

This is an exciting opportunity to join a pioneering team at the forefront of industrial and scientific innovation. As a key contributor, you will be instrumental in developing and deploying cutting-edge AI models that unlock profound insights from complex, real-world data. Your work will directly impact critical applications, driving advancements in areas like predictive maintenance, anomaly detection, and multi-modal sensor fusion, ultimately shaping the future of intelligent systems. Partnering with Aquent, you will contribute to a leading organization that is committed to pushing the boundaries of technology and making a tangible difference in various sectors.

**About the Opportunity**

Our partner is a global leader dedicated to solving some of the world's most complex challenges through technological innovation. They are building the next generation of intelligent systems, leveraging vast amounts of sensor, time series, and multi-modal data to create predictive and analytical solutions. This role offers a unique chance to work on large-scale, self-supervised "foundation" models, transforming raw data into actionable intelligence across diverse industrial and scientific applications. You will be part of a collaborative environment where your expertise in machine learning and data science will drive significant advancements and create lasting impact.

**What You Will Do**

As a core member of the team, you will be responsible for pioneering the development and deployment of advanced AI solutions, including:

Foundation Model Development: Architect, train, and deploy large-scale, self-supervised foundation models capable of learning rich representations from time series, sequential sensor data, textual, and vision data.

Model Fine-Tuning & Application: Adapt and fine-tune models for critical tasks such as anomaly/event detection, predictive maintenance, forecasting, classification, and multi-modal sensor fusion for industrial and scientific applications.

Advanced Data Processing: Implement sophisticated processing, augmentation, and feature engineering techniques for diverse financial, industrial, IoT, medical, and other sensor streams, handling both univariate and multivariate time series.

Sensor Data Expertise: Apply deep expertise in analyzing diverse sensor modalities (e.g., accelerometers, temperature, vibration, audio, images), managing varying sampling rates, synchronization, and real-world noise/artifact handling.

Multi-Modality Learning: Design and implement robust deep learning architectures that integrate heterogeneous data types (time series, images, text, audio, structured data) through cross-modal representation learning.

Cutting-Edge ML Techniques: Drive innovation using self-supervised and semi-supervised learning methods, including masked modeling, contrastive methods, temporal predictive coding, and multimodal alignment and fusion.

Model Architecture Exploration: Work with a wide array of model architectures, such as sequence models (RNNs, GRU/LSTM, TCN), 1D/2D/3D CNNs, Transformers, graph neural networks, diffusion/generative models, and multi-modal/fusion encoders.

Scalable Solutions: Implement transfer learning and fine-tuning strategies at scale, including prompt/adapter-based methods, temporal domain adaptation, and few-shot learning for specialized tasks.

Performance Evaluation: Define and apply comprehensive evaluation metrics for regression/classification (MSE, F1, AUC), time series similarity (DTW, correlation), event detection/segmentation (IoU, accuracy), and alignment with business/end-user KPIs.

High-Performance Software Development: Develop custom kernels and high-performance preprocessing solutions using Python (NumPy, SciPy, Pandas) and C++/CUDA.

Deep Learning Infrastructure: Leverage and optimize deep learning frameworks such as PyTorch (Lightning, Distributed), TensorFlow/Keras, and JAX/Flax for large-scale training on multi-GPU, multi-node clusters with mixed-precision and ZeRO optimization.

Robust Data Engineering: Build and maintain robust data pipelines for ingesting, cleaning, segmenting, and aligning large-scale, time-synchronized multi-sensor datasets.

Mathematical Foundations: Apply strong theoretical knowledge in Linear Algebra, Probability & Statistics, Optimization (stochastic, convex/non-convex, Bayesian), Signal Processing (Fourier/wavelet analysis, filters, resampling, noise modeling), and Numerical Methods (ODE/PDE solvers, inverse problems, regularization, time-frequency methods).

Collaborative Impact: Engage in cross-disciplinary teamwork with domain experts, engineers, product owners, and end-users, effectively communicating complex model behaviors, uncertainty quantification, and value impact.

**Must-Have Qualifications:**

* MS or Ph.D. in Computer Science, Data Science and AI, or a closely related quantitative field.

* Minimum of 3 years of relevant professional experience in data science and AI or related fields.

* Expert proficiency in Python, including libraries such as NumPy, SciPy, and Pandas.

* Demonstrated experience with leading deep learning frameworks like PyTorch (including Lightning and Distributed training), TensorFlow/Keras, or JAX/Flax.

* Proven ability to build, train, and deploy large-scale, self-supervised "foundation" models.

* Solid understanding and practical experience with self-supervised and semi-supervised learning techniques.

* Experience with various model architectures, including sequence models (RNNs, GRU/LSTM, TCN), CNNs, and Transformers.

* Proficiency in developing robust data pipelines for large-scale, time-synchronized multi-sensor datasets.

* Strong foundational knowledge in Linear Algebra, Probability & Statistics, and Optimization.

* Practical experience with Signal Processing techniques.

* Excellent collaboration and communication skills, with the ability to present complex technical concepts clearly to diverse audiences.

**Nice-to-Have Qualifications:**

* Experience with C++/CUDA for developing custom kernels and optimizing high-performance preprocessing.

* Familiarity with specific Transformer models such as BERT, ViT, or TimeSFormer.

* Knowledge of specific filters like Kalman or Savitzky-Golay in signal processing.

* Prior experience in industrial, IoT, medical, or financial sensor data domains.

* Experience with model interpretability, attention analysis, and uncertainty quantification.

#LI-LB2

#app