Satish Kumar

Key Skills

Software

Label Studio

Scale AI

OneForma

Top Subject Matter

Document Data Extraction

Quality Assurance

Driver Drowsiness Dataset

Top Data Types

Image

Text

Video

Document

Audio

Top Label Types

Prompt Response Writing SFT

Question Answering

RLHF

Text Generation

Translation Localization

Classification

Function Calling

Entity Ner Classification

Freelancer Overview

AI Trainer — Data Extraction QA, Voice AI & Behavioural Annotation (Alignerr - Freelance). Brings 3+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include OneForma and Outlier. AI-training focus includes data types such as Document, Video, and Audio and labeling workflows including Evaluation, Rating, and Classification.

Labeling Experience

Data Labeling

VideoObject Detection

• Classified driver drowsiness on a 4-level scale (Not Drowsy → Level 3: Very Drowsy/Asleep) from holistic video observation. • Labelled fine-grained eye behaviours: slow blinks, rapid blinks, prolonged closure (2–5s and 5s+), eye roll, squinting, straining to stay open. • Annotated head movement patterns: involuntary nodding, neck-muscle loss, head bobbing, shaking to stay alert, sustained head-lean-back. • Recorded body posture states and hand actions (eye/face rubbing, head-in-hands) relevant to fatigue detection. • Applied 'Cannot Determine' protocols rigorously when occlusion, blur, glare, or camera misalignment compromised reliable observation.

• Classified driver drowsiness on a 4-level scale (Not Drowsy → Level 3: Very Drowsy/Asleep) from holistic video observation. • Labelled fine-grained eye behaviours: slow blinks, rapid blinks, prolonged closure (2–5s and 5s+), eye roll, squinting, straining to stay open. • Annotated head movement patterns: involuntary nodding, neck-muscle loss, head bobbing, shaking to stay alert, sustained head-lean-back. • Recorded body posture states and hand actions (eye/face rubbing, head-in-hands) relevant to fatigue detection. • Applied 'Cannot Determine' protocols rigorously when occlusion, blur, glare, or camera misalignment compromised reliable observation.

2026 - Present

AI Trainer — Data Extraction QA, Voice AI & Behavioural Annotation (Alignerr - Freelance)

OneformaAudioFunction Calling

Trained voice AI models to recognize user intent and trigger correct function calls. Evaluated model responses for accurate API or function selection and parameterization. Provided structured failure-case feedback to improve AI reliability in real-world usage contexts. • Focused on real-world voice interaction tasks and intent recognition. • Rated performance of API/function selection and response correctness. • Supplied structured feedback for model retraining and error reduction. • Ensured robust and accurate function call behavior in simulated environments.

Trained voice AI models to recognize user intent and trigger correct function calls. Evaluated model responses for accurate API or function selection and parameterization. Provided structured failure-case feedback to improve AI reliability in real-world usage contexts. • Focused on real-world voice interaction tasks and intent recognition. • Rated performance of API/function selection and response correctness. • Supplied structured feedback for model retraining and error reduction. • Ensured robust and accurate function call behavior in simulated environments.

2025 - Present

AI Trainer — Data Extraction QA, Voice AI & Behavioural Annotation (Alignerr - Freelance)

OneformaDocument

Reviewed AI-extracted structured records from PDFs, emails, invoices, contracts, and scanned documents to validate data accuracy. Ensured correct extraction of fields such as dates, amounts, names, codes, and reference numbers, correcting errors and filling missed information. Flagged unclear or ambiguous data for escalation while maintaining thorough QA standards across various document schemas. • Worked with high-volume batches of mixed document types and changing annotation schemas. • Applied consistent QA protocols across multi-type document annotation projects. • Experienced with field-by-field comparison of AI output vs. source documents. • Adapted to rapid schema shifts and escalated particularly ambiguous records.

Reviewed AI-extracted structured records from PDFs, emails, invoices, contracts, and scanned documents to validate data accuracy. Ensured correct extraction of fields such as dates, amounts, names, codes, and reference numbers, correcting errors and filling missed information. Flagged unclear or ambiguous data for escalation while maintaining thorough QA standards across various document schemas. • Worked with high-volume batches of mixed document types and changing annotation schemas. • Applied consistent QA protocols across multi-type document annotation projects. • Experienced with field-by-field comparison of AI output vs. source documents. • Adapted to rapid schema shifts and escalated particularly ambiguous records.

2025 - Present

Ai traning

OneformaVideoObject DetectionClassification

Data Labeling • Classified driver drowsiness on a 4-level scale (Not Drowsy → Level 3: Very Drowsy/Asleep) from holistic video observation. • Labelled fine-grained eye behaviours: slow blinks, rapid blinks, prolonged closure (2–5s and 5s+), eye roll, squinting, straining to stay open. • Annotated head movement patterns: involuntary nodding, neck-muscle loss, head bobbing, shaking to stay alert, sustained head-lean-back. • Recorded body posture states and hand actions (eye/face rubbing, head-in-hands) relevant to fatigue detection. • Applied 'Cannot Determine' protocols rigorously when occlusion, blur, glare, or camera misalignment compromised reliable observation.

Data Labeling • Classified driver drowsiness on a 4-level scale (Not Drowsy → Level 3: Very Drowsy/Asleep) from holistic video observation. • Labelled fine-grained eye behaviours: slow blinks, rapid blinks, prolonged closure (2–5s and 5s+), eye roll, squinting, straining to stay open. • Annotated head movement patterns: involuntary nodding, neck-muscle loss, head bobbing, shaking to stay alert, sustained head-lean-back. • Recorded body posture states and hand actions (eye/face rubbing, head-in-hands) relevant to fatigue detection. • Applied 'Cannot Determine' protocols rigorously when occlusion, blur, glare, or camera misalignment compromised reliable observation.

2025 - Present

Hindi Language Data Annotator — OneForma (Freelance/Ongoing)

OneformaTextEntity Ner Classification

Participated in ongoing Hindi language AI annotation projects contributing labeled data and quality evaluations. Performed linguistic annotation, response evaluation, and dataset contribution tasks on proprietary platforms. Supported model evaluation and iterative training through structured data annotation and review. • Handled Hindi language NER and classification assignments. • Contributed to prompt evaluation and dataset improvement efforts. • Applied guideline-based linguistic annotation and accuracy checks. • Provided annotated data for continuous AI model refinement.

Participated in ongoing Hindi language AI annotation projects contributing labeled data and quality evaluations. Performed linguistic annotation, response evaluation, and dataset contribution tasks on proprietary platforms. Supported model evaluation and iterative training through structured data annotation and review. • Handled Hindi language NER and classification assignments. • Contributed to prompt evaluation and dataset improvement efforts. • Applied guideline-based linguistic annotation and accuracy checks. • Provided annotated data for continuous AI model refinement.

2024 - Present

Show more

Education

Work History