Feb 2025 – Present · 1 yr 1 mo · Remote

• 1. Led cross-functional pods as Pod Lead — owned delivery for Model improvement, coordinating engineers, ML researchers, and product managers to meet timelines and quality targets.
• 2. Managed project timelines, resources, and introduced process improvements that reduced fine-tuning cycle time by ~20% and improved release reliability.
• 3. Mentored and scaled engineering teams, establishing best practices for model evaluation, code reviews, and reproducible training pipelines to raise team velocity and quality.
• 4. Collaborated with stakeholders (PMs, data scientists, and infra) to translate business goals into technical roadmaps and production-ready AI features.
• 5. Designed and managed end-to-end LLM data pipelines (collection → preprocessing → augmentation) to ensure scalable datasets for enterprise deployments.
• 6. Specialized in fine-tuning large LLMs to improve model accuracy, robustness, and alignment, driving measurable product improvements.
• 7. Automate LLM training workflows using Python scripting, reducing manual intervention and streamlining iterative development processes.
• 8. Continuously evaluate and improve model performance through hyperparameter tuning, cross-validation, and advanced error analysis to ensure optimal outcomes.
• 9. Built and extended LLM applications with Python, PyTorch, Hugging Face, LangChain & LlamaIndex, tailoring behavior to client workflows and product needs.
• 10. Automated training, testing, and deployment pipelines, cutting manual effort and accelerating iteration cycles via CI-style workflows and Python tooling.
• 11. Implemented responsible AI practices — bias detection/mitigation and adversarial-reward checks that improved error-detection coverage and safer deployments.
• 12. Continually upskilled on GenAI technologies (RAG, multimodal LLMs, LangChain, LlamaIndex) and incorporated innovations into enterprise solutions

Oct 2024 – Jan 2025 · 3 mos · Remote

• 1. Specialized in fine-tuning large-scale language models (LLMs), including RLHF multimodality and SFT function calling tasks, delivering measurable improvements in model accuracy, reliability, and user alignment.
• 2. Built and enhanced LLM applications using Python, PyTorch, Hugging Face, and LangChain, tailoring models to client-specific workflows across multiple industries.
• 3. Optimized model performance through hyperparameter tuning, cross-validation, and error analysis, achieving improved response quality and reduced failure rates.
• 4. Automated training and evaluation workflows with Python scripting and custom pipelines, reducing manual effort and accelerating model iteration cycles.
• 5 Continuously upskilled on cutting-edge AI techniques (LangChain, LlamaIndex, multimodal LLMs, GenAI architectures), incorporating innovations into enterprise-level AI solutions.