Kriti Anandan
@kritianandan
Machine Learning Research Engineer at Skit.ai
Bangalore, India
Kriti Anandan is an experienced Machine Learning Research Engineer with expertise in developing advanced AI solutions, particularly in conversational AI and natural language processing. She has a strong background in building LLM-backed voicebots, implementing complex prioritization models, and optimizing speech recognition systems (VAD, Language ID). Her technical skills include Python, PyTorch, and AWS, demonstrated through research roles in speech emotion recognition and machine translation.
Experience
Machine Learning Research Engineer
Skit.ai
Research Intern (Master’s Thesis)
Speech Graphics
Investigated the effectiveness of label correction methods in improving the accuracy and interpretability of speech emotion recognition models.
Machine Learning Research Engineer
Skit.ai
Designed and built Skit’s first configurable LLM-backed voicebot for collections. Prototyped a prioritization model for debt collections. Conducted major experiments for the paper “Improving Spoken Language Identification with Map-Mix” (ICASSP’23). Developed processes to reduce label noise and built a prototype barge-in module for user interruptions. Increased intent F1 scores by 28% by adding a rejector model. Researched and deployed VAD models, achieving significant WER decreases for Tamil, Hindi, and English.
NLP Research Intern
LTRC, IIIT-Hyderabad
Developed algorithms to extract language patterns from NPTEL math video lectures for training data. Implemented CRF++ and deep learning models to detect mathematical expressions in NPTEL video transcripts, achieving a 71% F1 score.
Software Engineering Intern
Stride.ai
Built a text classification model for Arabic documents with an F1 of 80%. Created a model to detect underlined words using OpenCV. Developed modules to extract data points from unstructured text documents.
Education
The University of Edinburgh
MSc
Artificial Intelligence
Manipal Institute Of Technology
B.Tech
Computer Science Engineering