Amitesh Gangrade

WORK HISTORY

EDUCATION

ABOUT AMITESH GANGRADE

I’m a Machine Learning Engineer specializing in active learning, submodular optimization, and LLM-based agent systems. I’m currently pursuing my MS in Computer Engineering at UT Dallas, and I work as an ML Engineer Intern at Avawatz, where I build large-scale perception and data-centric AI systems for real-world environments.At Avawatz, I design end-to-end active learning pipelines for 3D object detection, leveraging SMI/SCG and GPU-accelerated submodular optimization to improve model accuracy and drastically reduce labeling costs. My work has produced 5%–14% rare-class AP improvements, 2%–4% mAP gains, and 30%+ reductions in annotation cost, scaling to 1M+ feature spaces under tight latency constraints.I’m also the co-author of a NeurIPS 2025 paper, where we introduce a unified framework for open-world object detection that boosts unknown object recall by 2.4× over state-of-the-art methods. My research interests center on data-centric ML, representation learning, autonomous perception, and open-world systems.Before transitioning into ML, I spent 3 years at HSBC as a Senior Software Engineer, modernizing cloud platforms, building high-scale ETL/data pipelines, and improving API architectures for global insurance products used by 500K+ customers. This gives me a strong engineering foundation and the ability to bring ML models all the way into production.I’m also an active open-source contributor. My projects include:mcp-google-email — merged into Anthropic’s MCP ecosystemOptimizIt — an optimization library supporting convex and submodular objectivesLLM agent tooling with ReAct, RAG, memory persistence, and structured tool invocationI love working on high-impact ML problems where systems engineering meets cutting-edge research, especially in:data quality, active learning, LLM agents, and scalable ML systems.If you\'re working on something exciting in ML, perception, or agentic systems — I’d love to connect.