Hi, I'm Tamzeed

Principal Applied Scientist, Oracle AI Labs (OCI)

Leading post-training efforts for LLM-based code and SQL generation.

Applied Scientist, Amazon AGI

Large Language Model fine-tuning, reinforcement learning, and LLM-based agent safety.

• Post-training science lead for Nimbus-AWS collaboartion for molecular property prediction. Work got highlighted at AWS reinvent'25 keynote speech.
• Research lead for GRPO training for domain-specific Python code generation using verifiable rewards.
• Led Nova model supervised fine-tuning for an MIT metamaterial design project featured in the AWS NYC Summit 2025 keynote.
• Designed data mixing experiments for PPO-based model customization to identify optimal data ratios across sources.
• Led end-to-end data generation and SFT to enhance Nova’s anomaly detection capabilities for time-series chart analysis.
• Developed and deployed an LLM-based safety model in production for browser agents (Amazon Nova Act).
• Engineered synthetic data generation pipelines for diverse safety scenarios and contributed to reward model training and evaluation for browser-agent RL.

Applied Scientist, Amazon Lab126

Developed multi-modal ML solutions for Echo devices and always-on audio sensing.

• Developed a multi-task DNN for Echo Show 10, combining speech source localization and VAD.
• Implemented an optimized sound event classification model for on-device DSP deployment.
• Led development of multi-device, speech-based orientation detection using multi-channel audio data.

Applied Scientist Intern, Amazon Lab126

Audio ML for smart devices.

• Developed deep neural networks for sound event detection and localization from multi-channel audio.

Summer Research Intern, Microsoft Research

Personalized audio experiences using sensor fusion.

• Developed ML models for HRTF personalization using integrated sensor fusion techniques.
• Mentored by Dr. Ivan Tashev.