TixelJobs
N
nanvia muse

AI ML Engineer – Computer Vision

KA, INPosted 3w ago
Computer VisionMid LevelFull-time

About the Role

AI/ML Engineer – Computer Vision

*Detailed job description \| Recruitment and hiring reference*

**Experience**

1–2 years preferred; candidates with \~1 year of strong hands\-on work considered

**Education**

B.E./B.Tech in CSE, AI/ML, ECE, IT or related discipline; M.Tech/M.Sc also welcome

**Reporting to**

Senior AI/ML Engineer / AI/ML Technical Lead, under CTO / Technical Director

**Indicative CTC**

₹5–8 LPA fixed target; flexible for demonstrated skills and budget approval

**Incentives**

Discretionary performance / delivery bonus, subject to approved company policy

**About Eigenstate AI**

Eigenstate AI is building a vision\-native Industrial Intelligence platform for manufacturing. We use computer vision, video analytics and edge AI to help factories understand production activity, throughput, utilization, process adherence and operational losses. The role involves practical work on real shop\-floor video data and customer pilots.

**Role purpose**

Build, test, fine\-tune and integrate computer vision capabilities under the guidance of the technical lead. This is an individual\-contributor engineering role, not a people\-management or independent platform\-architecture role.

**Key responsibilities**

· Implement computer vision features for object/person detection, tracking, segmentation, activity recognition and event counting.

· Evaluate and fine\-tune open\-source models (for example YOLO\-family architectures) for industrial video use cases.

· Develop Python/OpenCV preprocessing, inference, post\-processing and temporal event logic for production cycles and operator states.

· Prepare and curate datasets; coordinate annotation guidelines, train/validation splits and model experiments.

· Test model predictions against manually verified ground truth; calculate precision, recall, counting error and event\-level accuracy.

· Investigate false positives, missed events, occlusion, camera\-angle limitations and lighting changes; document root causes and fixes.

· Support RTSP/video ingestion, GPU inference, Linux deployment and integration with application APIs under team guidance.

· Write maintainable, modular, version\-controlled code, tests and technical documentation.

· Contribute reusable algorithms, labeled datasets, evaluation tools and domain\-specific logic to Eigenstate’s proprietary IP, respecting open\-source licenses.

· Collaborate with QA, annotation, application engineers and software developers to deliver pilot milestones.

Essential skills

· Hands\-on working experience with C\+\+ and a strong understanding of object\-oriented programming (OOP/OOPS) concepts.

· Good Python programming, debugging, data structures and Git fundamentals.

· Practical OpenCV and PyTorch experience; ability to run, modify, train and evaluate vision models.

· Understanding of detection, classification, tracking, image preprocessing and model evaluation.

· Experience handling images/videos, annotations and train/validation/test datasets.

· Comfort with Linux, basic command line tools, NumPy and common Python libraries.

· Ability to explain completed projects, trade\-offs, errors and lessons learned clearly.

**Good\-to\-have skills**

· YOLO, ByteTrack/DeepSORT, pose estimation, action recognition or temporal modeling.

· RTSP, FFmpeg/GStreamer, ONNX, TensorRT, CUDA or edge GPU deployment.

· Docker, REST APIs, SQL, basic MLOps / experiment tracking.

· Industrial automation, robotics, manufacturing or surveillance\-video analytics exposure.

**Education and experience**

Relevant bachelor’s degree or higher preferred. Practical projects, internships, GitHub repositories, deployed demos and demonstrable engineering capability will carry significant weight. Strong candidates with approximately one year of experience are welcome; a master’s degree is not mandatory.

**Performance expectations: first 90 days**

· Month 1: understand current pipelines, reproduce existing results and complete one small, reviewed fe
Share