Relevant Experience & Education Highlights

Expertise in optimizing large-scale generative AI models with transformers, diffusion models, and advanced inference techniques like FSDP and vLLM positions this candidate to excel in building scalable ML infrastructures for Adobe's Firefly generative AI initiatives.

1. Build and optimize infrastructures that power large foundation model training on thousands of GPUs

Proven track record in scaling AI training and inference systems with GPU-optimized techniques.

• Optimized large-scale Transformer and Diffusion models integrating FlashAttention, FSDP (Fully Sharded Data Parallel), vLLM, and speculative decoding for efficient inference and deployment.
• Led development of Retrieval Models leveraging Retrieval-Augmented Generation (RAG), Dense Retrieval (DR), Vector DBs including FAISS, Weaviate, Pinecone, and Hybrid Search (BM25, Neural).
• Developed self-improving AI architectures using continual learning, active learning, and adaptive fine-tuning pipelines.
• Advanced Knowledge Graph-based AI Reasoning by integrating structured data with transformer-based models.

2. Architect and optimize end-to-end ML pipelines, ensuring they're scalable, efficient, and robust

Hands-on experience architecting production-grade ML systems with MLOps tools and cloud services.

• Developed Generative Media solutions enhancing Text-to-Image (Stable Diffusion, DALL-E, Midjourney), Video Generation (Runway Gen-2, Sora), and Audio Generation (MusicLM, AudioCraft).
• Spearheaded architectures leveraging Mixture-of-Experts (MoE), Task-Specific Adapters, and Transformer-based models for scaling AI across domains.
• Utilized Apache Airflow, Apache Kafka, Amazon S3, Amazon Kinesis, and Amazon Redshift for data processing pipelines at a major enterprise.
• Deployed scalable systems with AWS services including Amazon EC2.

3. Dive deep into data to recommend the right models, evaluation metrics, and governance approaches

Deep domain knowledge in generative AI aligns with selecting optimal models for creative applications.

• Published and patented novel methodologies in memory-augmented reasoning, multimodal retrieval, and generative AI.
• Defined AI product roadmaps aligning state-of-the-art research with enterprise solutions for intelligent assistants and LLM-driven applications.
• Collaborated cross-functionally with leading research teams to integrate advanced AI memory reasoning, multimodal models, and retrieval-enhanced architectures into production.
• Drove innovation in recommendation and personalization systems using Deep Learning and Large Language Models (LLMs) at a luxury retail company.

Requirements & Candidate Alignment