At Ethara AI, I work on LLM fine-tuning and experimentation in Python, evaluating model outputs to improve response quality and pipeline reliability.
I contribute to RLHF and LLM post-training through response evaluation, ranking, and feedback generation. I also analyze model behavior to identify recurring failure patterns.
At Ethara AI, I lead a 20-member AI quality team, establishing evaluation standards and maintaining consistent quality benchmarks.
At PyVision Technologies, I worked on AI projects and designed an image-based burn analysis system. My projects include a multimodal pitch analyzer using Whisper and Librosa, and AI WasteGuard for detecting and classifying waste in water-body sonar images.
