Mithun Bachu

Analytical and detail-oriented AI Evaluation Specialist with over 2+ years of experience in AI data annotation, multimodal evaluation, quality assurance, RLHF workflows, and speech/video analysis. Experienced in evaluating AI-generated text, high-resolution image perception, audio, videos, and Computer Use Agent (CUA) interactions using structured rubric-based frameworks and Side-by-Side (SxS) methodologies. Proven track record in executing specialized multimodal annotation tasks including frame-level grounding, output comparison, and dense bounding box labeling.

Exceptional at identifying hallucinations, instruction-following failures, and temporal defects while maintaining strict quality and productivity targets in high-volume, secure remote environments.

AI Data Annotator & AI Evaluation Specialist at Outlier AI / Aether Projects (2024-02 – 2026-03)

Platforms & Tracking: Multimango System • Hubstaff Desktop App Integration

• Participated in Computer Use Agent (CUA) evaluation tasks involving browser workflows, UI understanding, and task completion verification.
• Evaluated AI-generated outputs across multimodal datasets, including complex image perception, captioning, and object recognition models.
• Performed grounding evaluation tasks involving dense bounding box annotation around complex UI elements in 1080p+ images, challenging referring expressions, image editing workflows, and multi-turn prompt evaluation.
• Engineered challenging testing instructions and descriptive context statements to evaluate and stress-test model object recognition boundaries.
• Verified and corrected model response trajectories through detailed output evaluation and comparison.
• Maintained a 95%+ accuracy score across 500+ annotations per week, ensuring high-quality training datasets while meeting strict productivity targets tracked via Hubstaff.
• Performed Side-by-Side (SxS) evaluations and rubric-based ranking tasks to identify model hallucinations and instruction-following failures.
• Evaluated UI screenshots, dashboards, forms, and browser workflows for grounding accuracy and interface understanding.
• Reviewed presentation slides and document layouts for structure, formatting consistency, and visual layout quality.

English & Telugu Transcription Specialist at Uber AI Solutions (2026-06 – Present)

• Transcribe English and Telugu audio following verbatim and clean-read transcription standards.
• Evaluate speech quality, transcript accuracy, grammatical syntax, and speaker consistency for processed assets.
• Assess speaker pronunciation, linguistic fluency, and overall speech clarity across multilingual datasets.

Telugu Voice Actor & AI Voice Training Contributor at Voices.com (2026-06 – Present)

Project: Casting Call: Voice Actors for AI Voice Training Projects – Telugu

• Record high-fidelity Telugu voice samples following detailed script instructions for speech synthesis training datasets.
• Deliver expressive, natural speech patterns across prompt-based recording assignments while adhering to acoustic rules.

Maps Personalization Contributor at Welocalize (2026-03 – 2026-06)

Projects: ARA MU | Arandee

• Evaluated map search queries and personalization outputs for relevance, geographic grounding, and search quality.
• Applied rigorous project guidelines to ensure consistency, accuracy, and quality across location-based datasets.

Telugu Voice Data Specialist at Lilt (2025-12 – 2026-06)

• Recorded natural Telugu voice samples to train automated speech recognition (ASR) systems following technical acoustic constraints.

Household Chores Video Data Contributor at Visora AI (2026-01 – 2026-04)

• Recorded targeted activity sequences adhering to computer vision dataset capture frames and camera tracking rules.

Video Data Contributor (First Person POV Videos) at Micro1 (2025-12 – 2026-03)

• Captured first-person point-of-view task sequences ensuring compliant spatial coverage and frame stability.

Multimodal Research & AI Platform Contributor at Prolific (2026-05 – Present)

• Completed research studies involving AI evaluation, user experience testing, surveys, and annotation tasks while maintaining high attention to detail and study compliance.
• Consistently surpassed data validation markers and passed stringent integrated attention-check metrics across distributed research sets.

Bachelor of Commerce (Computer Applications) in Computer Applications – Kakatiya University