Harshal J Hirpara

About

AI/LLM engineer who has shipped production systems end-to-end, including a note-taking product (Notey) combining audio/image capture, ASR, summarization, and a semantic chat agent over past notes. Also has applied ML experience in healthcare, collaborating directly with doctors to validate an EEG seizure-detection pipeline, and uses Kubernetes to optimize GPU usage for LLM training.

Experience

Education

Key Strengths

• Built and deployed a multi-component AI note-taking app (audio+images, ASR, summarization, semantic Q&A)
• Pragmatic debugging of ASR in noisy environments (switched from Whisper to AssemblyAI to improve robustness)
• Sets up transcription evals/benchmarks to maintain quality and reliability
• Production deployment with Docker and Fly.io
• Kubernetes-based GPU scaling/optimization for LLM training workloads
• Structured approach to model selection balancing context length, cross-vendor performance, and cost efficiency
• Designs agent workflows with LangChain/LangGraph patterns (ReAct/Reflect) and evaluates with open-source tools/custom metrics
• Effective collaboration with non-technical clinical stakeholders (weekly demos/validation with doctors for EEG seizure detection)

Reference Highlights

• Significant contributions as a founding engineer
• Able to ship stable and scalable production features under startup constraints
• Strong collaboration across teams
• Clear and prompt communication
• Breaks down complex technical ideas into simple terms
• Adapts communication style to the audience
• Good listener who asks the right questions
• Effective in cross-functional meetings and presentations
• Passion-driven
• Comfortable in fast to medium paced environments
• Improved output quality via custom search-augmented agent with tools
• Strong reverse engineering and QA/testing ability
• Potential/fit for team lead