Artificial Intelligence Engineer

Job Type: Full time

Role: AI/ML Engineer

Location: Hyderabad

Experience: 3+ years

Capabilities to design and develop as per the requirements

▪ Capabilities in tools to perform data preparation (Open CV, Kibana), Classification (Voxel51, Sklearn), Annotation (V7, Scripts), Power BI

▪ Capabilities to provide resources onshore/offshore modeland bench in case of additional needs

Item

Services Task – Name and Description

B1

Data Categorization –

• Analyse video data from all pilot installations to create an organized database categorized by environmental conditions including seasonal variations (summer, winter, fall, spring), lighting conditions (bright daylight, dim light, nighttime), and weather impacts (rain, snow, sunny).

• Document specific passenger interaction scenarios that pose challenges to the system, such as multiple passengers entering/exiting simultaneously, passengers carrying large objects, and cases that have triggered door operation issues or road calls.

• Create a statistical report analysing the frequency and patterns of different operational scenarios, identifying common patterns and potential system improvement areas

B2

Data Annotation –

• Perform precise annotations using the V7 tool to mark passenger positions and movements. This includes creating bounding boxes or segmentation masks around passengers, carried objects, and door areas according to provided guidelines.

• Conduct thorough quality assurance reviews of annotations completed by team members to ensure consistency and accuracy across the dataset.

• Maintain comprehensive log documenting edge cases, unusual scenarios, and situations where standard annotation guidelines may need clarification or modification.

B3

Comprehensive Testing

• Execute systematic testing of the current model using new video datasets collected from all pilot installations. Testing should cover various operational conditions and scenarios.

• Generate detailed daily performance reports including:

• Quantitative metrics: detection accuracy, precision, recall, and F1-scores

• Analysis of false positives and negatives for each specific scenario type

• System response time measurements for door operation decisions

• Performance variation analysis across different installs

B4

Issue Analysis

• Create detailed documentation of all detection failures and performance issues, including:

• Specific instances of missed passenger detections with timestamps and conditions

• Cases of incorrect door timing decisions and their potential causes

• Scenarios where system response time exceeded acceptable thresholds

• Analyze and document environmental and operational factors affecting model performance, creating a comprehensive issue categorization system

• • Develop detailed recommendations for model improvements based on observed failure patterns