Nishad Wajge
Researcher and engineer interested in the areas of software, game theory, statistics, and machine learning.
Best way to reach me is linkedin or alternatively by email: [firstname] dot [lastname] at gmail dot com
Research
Gradient Fidelity in Learned World Models
Standardized Difficulty Labels for Profiling LLMsShould Professionals Consider Their Adversary's Strategy?Experience
Amazon Lab126
Software Engineer Intern
- Shipped a C++ on-device speaker identification module to a production vehicle fleet, off the response path.
- Quantized the embedding model to INT8 ONNX for real-time recognition on ARM silicon, and gated impostor-prone utterances to raise the true accept rate.
- Built an MLOps loop fed by fleet data for continuous fine-tuning of SLMs, ASR and TTS, with encrypted voiceprints synced across devices.