Hello
I’m Harsha!
As a PhD student, I’m researching enchronic speech dynamics using self-supervised learning and I also work on bias discovery/mitigation in conversational AI.
I’m supervised by Éva Székely and Gustav Eje Henter. My research is also supported by the WASP program.
Current Research
I’m exploring enchronic speech dynamics using self-supervised learning and working on bias detection in conversational AI. Enchrony involves studying:
- How quickly we respond in conversations/Turn-taking.
- How we co-create meaning with our conversation partners.
- The social rules of conversation including slang, inside jokes etc.
- Other real-time interaction dynamics.
My goal is to uncover insights that advance our understanding of human communication and improve conversational AI systems.
Past Work
- Master’s in Electrical Engineering (Signal Processing): Built automatic speech recognition (ASR) models for literacy assessments in India, now in field use by an NGO. (2018-2021)
- MSc in Data Science & AI: Deepened my expertise in machine learning and AI. (2022-2024)
- Industry Experience:
- Developed speech assessment systems and voice activity detectors.
- Built LLM-based solutions for the aerospace industry, including a RAG-based system that automated repair scheme creation, saving repair engineers 98% of their writing time.
What Drives Me
I love tackling tough problems and finding unique solutions that push projects forward or take them to the next level.
General Interests
- Signal Processing
- Speech Technology
- Conversational AI
- Machine Learning
Feel free to hit me up!