Abhinav Tiwari

I’m an AI Data Specialist with 5+ years of experience working on real-world datasets used in speech recognition, LLMs, and multimodal AI systems.

A big part of my work involves video data collection, especially egocentric (first-person) recordings that simulate real human interactions and daily activities—something modern AI models rely heavily on.

Here’s what I offer:

Video Data Services:

Egocentric (first-person) video recording for AI training
Real-world scenario capture (household tasks, outdoor environments, public spaces, etc.)
Video game screen recording for dataset collection
Ensuring proper framing, lighting, motion consistency, and contextual clarity
Re-recording and refining videos based on QA feedback

Audio & Speech Services:

High-quality Hindi audio recording (clear pronunciation, tone, pacing)
Voice data aligned with visual or scripted scenarios
Consistent delivery across different contexts and datasets

Annotation & QA:

Multimodal annotation (text, audio, image, video)
Speech and dataset quality validation
Transcription review and dialogue evaluation
Linguistic and cultural accuracy checks (Hindi-focused)

LLM & AI Evaluation:

Prompt writing and response evaluation (Hindi & English)
RLHF-style feedback and output improvement
Identifying hallucinations, weak reasoning, and inconsistencies

I’ve contributed to multiple AI training pipelines where the focus is not just data collection, but making sure the data is actually usable and improves model performance.

What this really comes down to: I understand both sides—how data is created and how it’s evaluated. That helps me deliver work that doesn’t just pass guidelines, but actually adds value to the model.

I’m used to working with strict instructions, evolving project requirements, and tight deadlines while maintaining consistent quality.

If your project involves video datasets, Hindi audio, or multimodal AI training, I can handle it end-to-end—from recording to QA.