Machine Intelligence Analyst

Hello\nGreetings from ZettaMine Labs Pvt Ltd!!\n\nWe are looking for Machine Intelligence Analyst for PAN India.\n\nLooking only for Immediate Joiners\n\nJob Role : Machine Intelligence Analyst\nLocation : PAN India\nNotice Period : NA( As it is for Gig Workers)\nExperience : 2 to 4 years\nRelevant Exp : Minimum of 2 Years Experience in below list Skills\n\n'13; LLM internals (transformers, RLHF, attention)\n'13; Multi-modal AI systems\n'13; Adversarial ML & red-teaming\n'13; Inter-rater agreement metrics\n'13; NLP and computational linguistics\n'13; Safety evaluation & responsible AI\n'13; Prompt engineering & instruction evaluation\n'13; Python scripting & statistical analysis .\n\nMandatory :\nPhD in CS, ML, Computational Linguistics, Cognitive Science, or Mathematics (MS with exceptional industry/research experience considered)\n2–3+ years in AI/ML research, model evaluation, or applied NLP; demonstrated experience with LLM fine-tuning, alignment, or evaluation datasets\n\nJob Description:\nAs a Machine Intelligence Analyst, you will serve as a high-signal evaluator and adversarial tester for large language models and multi-modal AI systems. Working alongside frontier AI researchers and ML engineers, you will produce gold-standard evaluation data, surface failure modes, and stress-test model safety boundaries. Your expert judgment directly shapes the quality of AI systems used by the world’s leading technology organisations.\n\nKey Responsibilities:\n\nLLM Evaluation & Alignment\nConduct expert preference annotation and comparative evaluation for RLHF and RLAIF pipelines; assess outputs for accuracy, reasoning quality, coherence, and intent alignment\nGenerate and annotate SFT datasets; rate instruction-following quality, chain-of-thought correctness, and response calibration\n\nAdversarial Safety & Red-Teaming\nDesign and execute adversarial prompts to probe safety boundaries and jailbreak vectors; identify hallucination patterns, refusal evasion, and harmful output pathways\nContribute to safety evaluation rubrics, adversarial scenario libraries, and structured vulnerability documentation\n\nMulti-Modal Annotation & Quality\nAnnotate and evaluate text, image, audio, and video model outputs; apply domain knowledge (science, law, medicine, code, linguistics) to nuanced tasks\nUphold 98%+ accuracy; participate in inter-rater calibration; leverage Platina’s AI-assisted tooling to maximise throughput.\n\nPlease provide following detail along with your updated profile to\n\nPrimary Skill -\nTotal Experience -\nRelevant Experience :\nCurrent location\nHighest Qualificaiton:\nExpertise/Stream:\nDomain :\n\nI will be reaching out to you over a call or email as soon as possible.\n\nThanks & Regards,\nVyshnavi