Independent AI Trainer / Auditor (DataAnnotation, Telus, Lionbridge, Appen)
Hyderabad, India
ExpertData Annotation TechAppenLionbridge
Key Skills
Software
Data Annotation Tech
Appen
Lionbridge
Telus
Top Subject Matter
LLM Alignment
Model Safety & Localization
Healthcare AI
Top Data Types
Text
Image
Document
Top Task Types
RLHF
Bounding Box
Red Teaming
Evaluation Rating
Prompt Response Writing SFT
Object Detection
Freelancer Overview
• Enterprise Model Alignment (Metis PIG-35, 37, 39): Evaluated and compared two developmental model LLM responses, by crafting multi-turn prompts (in Hindi) as per specified categories, types, use cases, and sub-areas to align personalization targets, evaluating personal attributes and cultural nuances, writing respective model rationales and complex SXS overall quality evaluation in English.
• Health/Wellness Localization and Gemini apps alignment (Kardia SK-0220, HFA0304, HFA-0318) (Hindi-IN): Audited responses to the guided prompt on health conditions and images for clinical factuality on Gemini Live adhering strict medical guidelines. Generated multi-turn prompts with specified use cases and Compared responses of Gemili Live and Gemini Text (chat) for contextual drift, relevance and factuality to offer SXS overall quality evaluation in English.
• Foundational Technical Skills: Proficient in extracting Debug info, analyzing sian_profiles, and creating HTML verification files.
Expert
Labeling Experience
Google Search Quality Evaluator (Contractor via Appen/Lionbridge)
OtherText
I conducted high-stakes search engine quality evaluation for Google, focusing on Hindi and English search results related to YMYL (Your Money Your Life) content. My core tasks involved manual data annotation, query intent assessment, and detailed rating of web, image, and YouTube results by E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) and localized relevance guidelines. The role required consistent documentation, judgment of page quality (PQ), and measuring needs met (NM) for search users.
• Worked with search evaluation and annotation platforms for result classification and tagging.
• Evaluated cultural nuance, factual accuracy, and user safety in search and multimedia content.
• Applied rigorous quality and rating guidelines in both Hindi and English queries.
• Supported continuous improvement of Google’s core search and recommendations through annotation and judgment tasks.
I conducted high-stakes search engine quality evaluation for Google, focusing on Hindi and English search results related to YMYL (Your Money Your Life) content. My core tasks involved manual data annotation, query intent assessment, and detailed rating of web, image, and YouTube results by E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) and localized relevance guidelines. The role required consistent documentation, judgment of page quality (PQ), and measuring needs met (NM) for search users.
• Worked with search evaluation and annotation platforms for result classification and tagging.
• Evaluated cultural nuance, factual accuracy, and user safety in search and multimedia content.
• Applied rigorous quality and rating guidelines in both Hindi and English queries.
• Supported continuous improvement of Google’s core search and recommendations through annotation and judgment tasks.
2019 - Present
Independent AI Trainer (Telus International)
OtherTextBounding Box
Labeling of Hindi and Bengali letters and word images as per guidelines of Telus International.
Labeling of object images as per Telus/Lionbridge AI guidelines
Labeling of Hindi and Bengali letters and word images as per guidelines of Telus International.
Labeling of object images as per Telus/Lionbridge AI guidelines
2020 - 2021
Human movement recognition
VideoObject Detection
Hand Movements Project.
Summary
The task consists of taking 4 short videos of 30 seconds each per participant (project is handed out in pair). The pair have to upload 8 videos of 30 seconds each in the end. 4 from each partner in different scenarios.
Person 1 is: SADHAN GHOSH with REFERENCE NUMBER: 17511359
Person 2 is: KOUSTAV GHOSH with REFERENCE NUMBER: 17511359_b
For this task you will need the following:
· A camera with a minimal resolution of 12 MP (iPhone 7 or higher or any other phone device with a 12 MP camera)
· Good internet connection for uploading
·
Detailed instructions of basic task
Each 30-second videos must be divided as followed:
4 seconds of finger gestures. Example: tapping, pinch, spreading, typing, pulling, grasp etc…
4 seconds of clapping hands
4 seconds of moving your hands in front of the camera by waving and bring them close and far from the camera
4 seconds of rubbing your hands against each other
4 seconds of occluding one hand partially or fully using the other hand or by moving it outside the view
10 seconds of improvisation with your hands. Creativity is encouraged. Examples: make fists, rotate your hand etc… Note that each improvised gesture should last 2 to 4 seconds. So improvise 3 to 4 gestures in that 10-second segment.
Video scenarios and conditions for person 1
· Video 1 taken inside in bright lighting mode. Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned above your hands. Second person’s hand(s) showing from time to time during the video. Please have some distance from the camera, so that the forearms are visible. Please refer to the example video “Two people hands example” below.
· Video 2 taken inside in low lighting mode (but still visible). Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned on the side of the hands (left or right side of the hands, forearms still to be visible so keep an adequate distance from the camera). Second person’s hand(s) showing from time to time during the video. Please refer to the example videos below.
· Video 3 taken outside in daylight. Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned in front of the hands (facing the hands). Only one person hands are visible during the whole video. Please have some distance from the camera, at least the person to be visible until the neck(no faces or the video will be failed). Please refer to the example videos.
· Video 4 taken outside in daylight. Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned above your hands. Second person’s hand(s) showing from time to time during the video. Please have some distance from the camera, so that the forearms are visible Please refer to the example videos.
·
Video scenarios and conditions for person 2
· Video 1 taken inside in bright lighting mode. Bare hands, bare feet and camera positioned on top of the hands. Second person’s hand(s) showing from time to time during the video . Please have some distance from the camera, so that the forearms are visible as well as the bare feet. Please refer to the example videos.
· Video 2 taken inside in low lighting mode (but still visible). Bare hands and camera positioned in front of the hands (facing the hands). Second person’s hand(s) showing from time to time during the video. Please have some distance from the camera, at least the person to be visible until the neck (no faces or the video will be failed).
· Video 3 taken inside in low lighting mode (but still visible). Holding something in one or both hands (pen, ball, keys, cell phone, etc.), bare feet and camera positioned above the hands. Only one person hands are visible during the whole video. Please have some distance from the camera, so that the forearms are visible as well as the bare feet. Please refer to the example videos.
· Video 4 taken outside in shadow or low lighting mode (but still visible). Bare hands and camera positioned near person’s 2 nose (nose view). The person recording has to hold the phone close to the recorded person’s nose facing the hands, make sure forearms are visible too. Only one person hands are visible during the whole video. Please refer to the example videos “Nose View”.
Hand Movements Project.
Summary
The task consists of taking 4 short videos of 30 seconds each per participant (project is handed out in pair). The pair have to upload 8 videos of 30 seconds each in the end. 4 from each partner in different scenarios.
Person 1 is: SADHAN GHOSH with REFERENCE NUMBER: 17511359
Person 2 is: KOUSTAV GHOSH with REFERENCE NUMBER: 17511359_b
For this task you will need the following:
· A camera with a minimal resolution of 12 MP (iPhone 7 or higher or any other phone device with a 12 MP camera)
· Good internet connection for uploading
·
Detailed instructions of basic task
Each 30-second videos must be divided as followed:
4 seconds of finger gestures. Example: tapping, pinch, spreading, typing, pulling, grasp etc…
4 seconds of clapping hands
4 seconds of moving your hands in front of the camera by waving and bring them close and far from the camera
4 seconds of rubbing your hands against each other
4 seconds of occluding one hand partially or fully using the other hand or by moving it outside the view
10 seconds of improvisation with your hands. Creativity is encouraged. Examples: make fists, rotate your hand etc… Note that each improvised gesture should last 2 to 4 seconds. So improvise 3 to 4 gestures in that 10-second segment.
Video scenarios and conditions for person 1
· Video 1 taken inside in bright lighting mode. Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned above your hands. Second person’s hand(s) showing from time to time during the video. Please have some distance from the camera, so that the forearms are visible. Please refer to the example video “Two people hands example” below.
· Video 2 taken inside in low lighting mode (but still visible). Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned on the side of the hands (left or right side of the hands, forearms still to be visible so keep an adequate distance from the camera). Second person’s hand(s) showing from time to time during the video. Please refer to the example videos below.
· Video 3 taken outside in daylight. Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned in front of the hands (facing the hands). Only one person hands are visible during the whole video. Please have some distance from the camera, at least the person to be visible until the neck(no faces or the video will be failed). Please refer to the example videos.
· Video 4 taken outside in daylight. Holding something in one or both hands (pen, ball, keys, cell phone, etc.) and camera positioned above your hands. Second person’s hand(s) showing from time to time during the video. Please have some distance from the camera, so that the forearms are visible Please refer to the example videos.
·
Video scenarios and conditions for person 2
· Video 1 taken inside in bright lighting mode. Bare hands, bare feet and camera positioned on top of the hands. Second person’s hand(s) showing from time to time during the video . Please have some distance from the camera, so that the forearms are visible as well as the bare feet. Please refer to the example videos.
· Video 2 taken inside in low lighting mode (but still visible). Bare hands and camera positioned in front of the hands (facing the hands). Second person’s hand(s) showing from time to time during the video. Please have some distance from the camera, at least the person to be visible until the neck (no faces or the video will be failed).
· Video 3 taken inside in low lighting mode (but still visible). Holding something in one or both hands (pen, ball, keys, cell phone, etc.), bare feet and camera positioned above the hands. Only one person hands are visible during the whole video. Please have some distance from the camera, so that the forearms are visible as well as the bare feet. Please refer to the example videos.
· Video 4 taken outside in shadow or low lighting mode (but still visible). Bare hands and camera positioned near person’s 2 nose (nose view). The person recording has to hold the phone close to the recorded person’s nose facing the hands, make sure forearms are visible too. Only one person hands are visible during the whole video. Please refer to the example videos “Nose View”.
2020 - 2020
Education
S
State Council for Engineering & Technical Education, Government of West Bengal
Diploma in Electrical Engineering, Electrical Engineering
