Miss Parshati Saha

AI Content & Safety Specialist with roots in academic writing, now shaping safer, more intuitive AI through LLM evaluation, prompt optimization, safety alignment, and Supervised Fine-Tuning (SFT), including multimodal testing (image, audio, video). Experienced in building high-accuracy datasets and ethical frameworks via collaborations with global teams, including projects with NVIDIA and a major Big Tech AI initiative. Alongside, I've explored freelancing, applying this craft across diverse, real-world contexts, including Red teaming and mitigation strategy development.

My approach blends analytical precision with genuine curiosity about how people interact with technology, with a focus on conversational safety, cultural nuance, persona alignment, and RLHF evaluation.

Freelance AI Content & Safety Specialist - Self-employed - Remote

(2026-01)

• Executed independent projects in LLM evaluation, prompt design, safety alignment, and Supervised Fine-Tuning (SFT) across varied domains.
• Conducted Red Teaming to identify vulnerabilities and edge cases, applying mitigation strategies to strengthen model robustness and safety guardrails.

Data Analyst - Innodata Inc. - Remote

(2024-07 - 2025-08)

Evaluated AI outputs for accuracy, bias, tone, safety, and instruction adherence across workflows. Built mitigation datasets and classified prompts to strengthen safety enforcement.

• Assessed multimodal outputs (text, image, audio, video) using structured metrics for relevance, cultural nuance, and truthfulness.
• Refined persona alignment based on user intent, demographics, and engagement patterns.
• Analyzed search intent, measuring coherence, utility, and harmfulness via the Three Hs framework.
• Contributed across RLHF pipelines, MASE workflows, plugin ecosystems, and comparative scoring.
• Conducted text-to-image and visual evaluations for faithfulness and interpretative accuracy.
• Performed formatting and LaTeX structuring for technical content accuracy and consistency.
• Moderated NSFW content, applying nuanced judgment to ensure policy compliance.
• Delivered qualitative and quantitative evaluations to support safe, reliable AI deployment.

Associate Prompt Engineer - Spectrum Consultants India Pvt. Ltd. - Remote

(2023-09 - 2024-05)

Designed prompt frameworks, personas, and multilingual interaction flows to elevate AI responsiveness. Partnered with TTS and NLP teams to refine NLG and NLU performance.

• Developed narrative prompts for storytelling, captioning, and multimodal generation.
• Contributed to NVIDIA's RTX AI initiatives, supporting evaluation of AI-assisted rendering outputs, including systems leveraging real-time ray tracing and multimodal inputs.
• Contributed linguistic insights to the NeMo project, enhancing contextual precision and tonal fluency.
• Optimized dialogue systems through flow analysis, feedback loops, and safety tuning.

Academic Content Writer - Assurex - Kalyani

(2023-04 - 2023-06)

Produced research-driven academic and industry content with structural and linguistic finesse. Edited for clarity, accuracy, and publication readiness. Curated sources to deepen narrative quality and intellectual rigor.

B.A - English Honours - University of Calcutta (2015 - 2018)

12th Standard - BDM International (2013 - 2015)

Matriculation - BDM International (2013)