Position: Software Engineer / Senior Software Engineer, ML & ADAS Perception

Years of Experience: 2-3 Years

Location: Belagavi/Bangalore

Job Brief:

The world of ADAS and AV is undergoing a sea-change. OEMs, tier-1s, and regulators are all demanding measurable and quantifiable metrics of ADAS/AV safety. Data-driven simulation-based validation of ADAS/AV stacks is fast becoming the cornerstone of generating and verifying these metrics. Tools and Technologies based on the concept of Data-to-Simulation are driving this simulation-based validation approach. Vayavya is working on developing such tools, where every vehicle, pedestrian and cyclist in the recorded videos and sensor data are enumerated and tracked and every actor trajectory is located accurately on a world digital twin that can then be played in any simulator of choice.

Job Description:

Tool Development

•  Train and tune learning models to extract actor behaviour from a set of recorded ADAS/AV data
• Apply ADAS tracking and filtering algorithms to aid the learning models
• Build an automated simulation artefact generation ML pipeline.

Validation

• Check that what the pipeline produces is correct
• Build the tests, metrics, and review tools that measure each stage of the pipeline, and feed the corrections back into model training.

Technical Skills:

• Good programming skills in Python. C++ is a plus.
• Hands-on with a deep learning framework, PyTorch preferred.
• Working knowledge of computer vision: object detection, multi-object tracking, segmentation, depth estimation.
•  Understanding of camera geometry: intrinsics, extrinsics, lens distortion and projection.
• Training and fine-tuning models: datasets, labelling, evaluation metrics.
• Comfortable with linear algebra, probability, and basic state estimation such as Kalman filters.
• Exposure to Linux OS, Linux tools, and Git.

Good to Have (Non-mandatory):

•  Handling large datasets of video and logs with NumPy, pandas and plotting tools.
•  Writing test suites and regression checks, using pytest or similar.
• Understanding of evaluation metrics for detection and tracking: precision, recall, IoU, and tracking scores such as MOTA (Multiple Object Tracking Accuracy) or HOTA (Higher Order Tracking Accuracy).
• Exposure to ADAS simulation standards like ASAM OpenDRIVE, ASAM  OpenSCENARIO, and simulators like esmini or CARLA

Non-Technical Skills:

• Strong analytical and problem-solving skills
•  Good debugging skills
•  Excellent verbal and written communication skills
• Self-managed and able to learn and adapt