Shivjot

Multilingual AI Data Specialist and Project Coordinator with 2+ years of experience leading quality operations and coordinating multilingual teams across SFT, RLHF, red-teaming, and multimodal model evaluation projects for AI labs and data platforms including xAI, Labelbox/Alignerr, Invisible Technologies, and Outlier AI/Scale AI. Proven track record coordinating day-to-day operations, monitoring team and individual performance, and implementing quality improvement plans as a Project Coordinator and Team Lead. Currently focused on AI model refinement and voice/audio quality at xAI.

AI Tutor (Multilingual) – AI Model Refinement & Voice Quality at xAI (2026-05 – Present)

• Refine large language model outputs across reasoning, conversational quality, and instruction-following, with a primary focus on audio and voice-based responses, including Grok's voice systems.
• Evaluate and correct voice output for naturalness, pronunciation, tone, and linguistic accuracy across English, Hindi, and Punjabi.
• Identify model errors in spoken and conversational responses and provide structured feedback to drive iterative model improvements.
• Collaborate with cross-functional teams to improve multilingual voice and conversational AI performance.

Project Coordinator / Team Lead (English & Hindi) – Multimodal AI Red Teaming & Quality Operations at Labelbox / Alignerr (2025-09 – 2025-11)

• Oversaw day-to-day operations across labeling and review workflows, monitoring individual and team performance against quality and productivity targets.
• Led scheduled calls and office hours to resolve annotator queries, gather feedback, and identify process gaps.
• Tracked productivity and quality metrics and implemented improvement plans to maintain consistent standards.
• Delivered projects on schedule while maintaining strong cross-team communication across multilingual AI data initiatives.

Advanced AI Data Annotator (Bilingual) & Reviewer – Hindi & English Squad at Labelbox / Alignerr (2025-03 – 2026-05)

• Annotate and review multilingual AI training data, progressing from labeling to senior review responsibilities.
• Designed multi-turn conversations under defined constraints for Verifiable Instruction Following (VIF) projects to test model compliance and failure modes.
• Performed high-accuracy transcription of human–LLM audio interactions and supported multimodal red-teaming across text, image, and video.
• Contributed to SFT, RLHF, RAG, and community-alignment projects, including India-specific cultural and tradition-based evaluations.

AI Data Trainer (English & Hindi) – Model Training, Evaluation & Linguistic Quality at Invisible Technologies (2025-02 – 2025-09)

• Contributed to Google's Project MAGI, including the HCAI initiative, evaluating user queries against multiple model-generated outputs.
• Selected higher-quality responses based on dynamic, project-specific criteria including relevance, correctness, clarity, and alignment with evolving requirements.
• Maintained consistent, high-quality judgments across shifting evaluation criteria to support large-scale model improvement.

Gen AI Data Trainer (English, Hindi & Punjabi) | L10 Reviewer | Quality Control Manager at Remotasks / Outlier AI / Scale AI (2024-01 – 2025-01)

• Designed single- and multi-turn prompt-response pairs for Supervised Fine-Tuning (SFT) projects.
• Evaluated and rated model outputs for quality, safety, and instruction-following in RLHF and Safety RLHF workflows.
• Performed span-level factuality verification with credible source citations.
• Assessed voice-based model responses against defined linguistic and performance metrics.

Honours Bachelor of Information Sciences in Cyber Security – Sheridan College