- Company Name
- Mitsubishi Electric Research Laboratories
- Job Title
- Internship - Sensor Reasoning Models
- Job Description
-
**Job title**: Internship – Sensor Reasoning Models
**Role Summary**:
Conduct foundational research on sensor reasoning models that integrate multimodal perception (RF, infrared, LiDAR, event camera) with higher‑level text, visual, and multimodal reasoning. Develop and evaluate algorithms, design experiments on in‑house testbeds, and prepare scholarly outputs (publications, patents).
**Expectations**:
* 3‑month internship, flexible start from October 2025.
* Deliver research progress reports, algorithm prototypes, benchmark results, and draft publications/patents.
* Collaborate closely with senior researchers.
**Key Responsibilities**:
1. Research and develop algorithms that bridge perception (detection, segmentation, tracking) with reasoning over sensor streams.
2. Integrate LLMs/VLMs with multimodal sensor outputs (point clouds, radar heatmaps, BEV features).
3. Design and run experiments on in‑house testbeds; construct reasoning‑centric benchmarks (QA, temporal prediction).
4. Manage and analyze large multi‑sensor datasets (nuScenes, Waymo, Argoverse, MMVR, HIBER, RT‑Pose, K‑Radar).
5. Prepare results for publication in top venues and contribute to patent submissions.
6. Maintain reproducible code, scalable data pipelines, and GPU cluster job scheduling.
**Required Skills**:
* Proven experience in text, visual, and multimodal reasoning (e.g., VQA, temporal/spatio‑temporal reasoning, chain‑of‑thought).
* Expertise aligning or conditioning LLMs/VLMs on sensor outputs (point clouds, radar).
* Strong foundation in state‑of‑the‑art transformer and diffusion perception models (DETR, DiffusionDet).
* Hands‑on with large multi‑sensor datasets and ability to create custom benchmarks.
* Proficiency in Python and deep‑learning frameworks (PyTorch, JAX); knowledge of GPU cluster scheduling and data‑pipeline scaling.
* Publication record in top-tier venues (CVPR, ICCV, ECCV, NeurIPS, ICLR, ICML).
* Understanding of sensor fundamentals (RF, infrared, LiDAR, event cameras); radar knowledge (FMCW, MIMO, Doppler signatures, point clouds).
* Awareness of recent radar perception research (TempoRadar, SIRA, MMVR, RETR).
**Required Education & Certifications**:
* Advanced degree (MSc or PhD) in Computer Vision, Machine Learning, Robotics, or related field.
* Coursework or research experience in deep learning, multimodal fusion, and sensor perception.