Vipul S.

Key Skills

Software

No software listed

Top Subject Matter

Chemistry

Physics

Maths

Polity

Economics

Top Data Types

Image

Video

Text

Top Task Types

Point/Key Point

Text Generation

Classification

Question Answering

Entity (NER) Classification

Segmentation

Freelancer Overview

During my time at SoftAge.ai, I have gained hands-on experience in the training, evaluation, and fine-tuning of experimental Large Action Models (LAMs). My work primarily revolves around designing and engineering diverse, high-quality adversarial prompts and complex instruction datasets to rigorously test model capabilities across various domains. I actively analyze and review AI-generated response pairs, evaluating them against strict parameters such as factual correctness, instruction-following, logical reasoning, and helpfulness to isolate flaws and eliminate model hallucinations. Beyond standard data labeling and reinforcement learning through human feedback (RLHF), I bridge the gap between AI annotation and product engineering. I design and develop the internal learning portals used to streamline these training workflows using modern frontend frameworks like Next.js.

Labeling Experience

Education

Work History