Machine Learning Engineer/Data Scientist (Trajectory & Behavior Prediction)

Experience : 5+ Years

Industry : Autonomous Vehicles / Intelligent Transportation Systems (ITS)

Focus : Probabilistic Forecasting, MARL, and Risk Scoring

Role Summary :

We are seeking a high-caliber Machine Learning Engineer / Data Scientist to architect the next generation of predictive models for autonomous behavior and traffic safety. In this role, you will be responsible for designing high-fidelity models that forecast trajectories and assess crossing probabilities for diverse traffic participants. You will bridge the gap between simulation and real-world deployment by implementing Multi-Agent Reinforcement Learning (MARL) frameworks and time-shifted risk scoring mechanisms.

This position requires a deep technical understanding of Vehicle Kinematics and the ability to build sophisticated scoring algorithms based on multi-dimensional data severity. Working closely with simulation and data engineering squads, you will ensure that streaming feature pipelines translate into accurate, event-level predictions that drive cooperative vehicle behaviors.

Responsibilities :

- Predictive Trajectory Modeling : Design and deploy deep learning models (e.g., LSTMs, Transformers, or Graph Neural Networks) to forecast participant trajectories and behavior patterns in complex traffic environments.

- Risk Scoring Architecture : Develop sophisticated risk scoring mechanisms using Time-Shifted Risk Prediction and sliding time windows to identify potential hazards before they materialize.

- MARL Framework Implementation : Architect and train Multi-Agent Reinforcement Learning (MARL) frameworks to simulate cooperative and competitive behaviors among autonomous agents.

- Simulation & Ground Truth Integration : Partner with simulation teams to integrate high-fidelity ground truth scenarios and curate replayable datasets for continuous model training.

- Multi-Dimensional Scoring : Build and refine scoring algorithms that weight different data dimensions based on the severity and impact of predicted events.

- Advanced Performance Evaluation : Rigorously evaluate model efficacy using precision-recall curves, F1-scores, and specific event-level accuracy metrics tailored for safety-critical systems.

- Feature Pipeline Engineering : Collaborate with data engineers to define low-latency feature pipelines and manage high-throughput streaming inputs for real-time inference.

- Probabilistic Forecasting : Implement Bayesian models or probabilistic forecasting methods to quantify uncertainty in crossing probabilities and participant intentions.

- Behavioral Simulation : Utilize replayable datasets to validate model performance against historical edge cases and diverse traffic scenarios.

Technical Requirements :

- Applied Data Science : 5+ years of experience in data science, with a proven track record in real-time systems or simulation-based environments.

- Deep Learning Frameworks : Expert-level proficiency in Python and advanced libraries including PyTorch or TensorFlow, NumPy, and Pandas.

- Probabilistic Modeling : Strong experience in Time-Series Analysis, Bayesian inference, and probabilistic forecasting.

- Reinforcement Learning : Hands-on understanding of Reinforcement Learning (RL), with specific exposure to multi-agent settings and cooperative behavior training.

- Domain Knowledge : Functional understanding of Vehicle Kinematics, intelligent transportation systems (ITS), and spatial-temporal data structures.

Preferred Skills :

- Spatial-Temporal Analysis : Experience working with Graph Neural Networks (GNNs) for modeling interactions between multiple traffic participants.

- Cloud & Edge Deployment : Familiarity with deploying ML models at the edge or within containerized cloud environments for real-time inference.

- Kinematic Constraints : Ability to incorporate physical vehicle constraints and non-holonomic motion models into predictive frameworks.

- Stream Processing : Understanding of streaming data platforms (e.g., Kafka, Flink) to facilitate real-time feature engineering.

- Communication Mastery : Ability to articulate complex probabilistic results and risk scores to cross-functional stakeholders in simulation and hardware teams.