Looking for a job

Ayush Vishwakarma

@ayushvishwakarma2

Data analyst working on AI evaluation at Outlier.ai and Turing; built a Python tool to benchmark model response quality.

India

Message

What I'm looking for

I’m looking to build on my data-cleaning, validation, and AI evaluation experience with Power BI, creating interactive dashboards and data-driven business insights.

At Outlier.ai and Turing, I train and evaluate large language model outputs, applying structured rubrics and delivering preference labels across coding, reasoning, and general-knowledge tasks. I also document model failure cases with severity ratings for downstream fine-tuning.

I built a Python benchmarking tool to score labeled responses using BLEU, ROUGE, and a custom factual-accuracy metric. I’ve also contributed to Label Studio and built OffSOS, an Android app for offline emergency coordination over mesh networking.

Experience

Work history, roles, and key accomplishments

Current

AI Model Trainer

Current

Outlier.ai & Turing

Apr 2026 - Present (6 months)

Trained and evaluated large language model outputs by ranking responses, writing and refining prompts, and applying structured rubrics for factual accuracy, coherence, and safety. Delivered RLHF-style preference labels across coding, reasoning, and general-knowledge tasks.

LLM Evaluation Prompt Engineering RLHF Data Annotation Rubric Based Grading Python

AI Developer & LLM Evaluator

Independent / Freelance

May 2025 - Dec 2025 (7 months)

Benchmarked GPT-4o, Gemini 1.5 Pro, and Claude Sonnet across hundreds of labeled response pairs using BLEU, ROUGE, and a custom factual-accuracy metric. Evaluated model outputs for factual accuracy, coherence, and safety, applying structured rubrics and documenting findings.

LLM Benchmarking BLEU ROUGE Factual Accuracy Python Data Analysis