Manoj Kumar

AI/ML Professional | Generative AI & Speech Analytics Expert

Building Intelligent Agents for Business Transformation

Leveraging 4+ years of expertise in Generative AI, Speech & Audio Analytics, and NLP to create scalable, cloud-ready AI solutions that drive real-world impact.

Let's Connect

About Me

As an AI/ML professional with around 4 years of hands-on experience, I specialize in Speech & Audio Analytics, Generative AI, and Natural Language Processing (NLP). My passion lies in developing and deploying cutting-edge solutions, from Speech-to-Text (STT) and Text-to-Speech (TTS) pipelines to integrating Large Language Models (LLMs) for real-time applications.

I am the architect behind "iSTENO", a groundbreaking speech analytics tool that significantly reduced contact center wrap time by 50%. With a strong foundation in Python, Machine Learning, and Deep Learning, I am dedicated to building robust, scalable, and cloud-ready AI solutions that solve complex business challenges.

Education

Technical Skills

Programming Languages

  • Python (NumPy, pandas, scikit-learn, TensorFlow)

Machine Learning/Deep Learning/Deployment

  • Model building (classification, regression, clustering)
  • Model Deployment

Natural Language Processing (NLP) Tools

  • NLTK, spaCy, Transformers
  • Hugging Face Transformers, BERT, GPT, T5

Data Visualization

  • Matplotlib, Seaborn

Version Control

  • Git

Areas of Expertise

  • Data Science, Statistics
  • Natural Language Processing (NLP)
  • Generative AI / Large Language Models

Professional Experience

Manager: Business Consulting - Speech AI and GenAI Development

HSBC (Innovation Team) | Aug '21 - Present

  • Led the development of iSTENO, an advanced GenAI-powered stenography system with Speech-to-Text (ASR), TTS, Summarization, and Sentiment Analysis capabilities. Integrated models inspired by Whisper, Google speech to text and transformer-based summarization, reducing call wrap-up time by 50%.
  • Fine-tuning ASR/TTS Pipelines: Enhanced ASR performance using domain-specific audio data and semantic corrections. Tuned TTS output for emotional consistency in playback, achieving 85% transcription and playback accuracy in production.
  • Fraud Call Detection (NLP-based): Developed a fraud classification engine using NER and semantic embeddings to identify fraudulent customer calls with 70% accuracy on real datasets.
  • Data Quality Automation: Designed NLP workflows to correct grammatical and spelling errors in transcriptions, improving data fidelity for analytics modules.
  • Call Quality Monitoring System: Built a rule-based and AI-assisted quality monitoring module that evaluates agent-customer conversations against standard operating procedures (SOPs). Used NLP-based intent recognition, custom keyword spotting, and compliance tracking to flag missed greetings, disclosures, and closing statements. Helped reduce manual QA effort by 65% and improved procedural compliance scores across teams.

Projects & Achievements

iSTENO – Speech Analytics Platform

Led the development of iSTENO, an advanced GenAI-powered stenography system with Speech-to-Text (ASR), TTS, Summarization, and Sentiment Analysis capabilities. Integrated models inspired by Whisper, Google speech to text and transformer-based summarization, reducing call wrap-up time by 50%.

Fraud Call Detection (NLP-based)

Developed a fraud classification engine using NER and semantic embeddings to identify fraudulent customer calls with 70% accuracy on real datasets.

Call Quality Monitoring System

Built a rule-based and AI-assisted quality monitoring module that evaluates agent-customer conversations against standard operating procedures (SOPs). Used NLP-based intent recognition, custom keyword spotting, and compliance tracking to flag missed greetings, disclosures, and closing statements. Helped reduce manual QA effort by 65% and improved procedural compliance scores across teams.

PyPI Libraries

Hack the Hacker, HSBC Participant

Built a machine learning model for fraud detection, achieving a PR AUC score of 0.31. This showcases my strong skills in building predictive models.

AI vs. Human Essay Detection (7th Place)

Secured 7th place in a competition identifying AI-written essays. This experience honed my expertise in Natural Language Processing (NLP) and machine learning.

Published Author

Co-authored a research paper titled "Forecasting Financial Fraud Through Machine Learning: An Indian Perspective." This demonstrates my research abilities and interest in financial applications of machine learning.

Get in Touch

I'm always open to discussing new projects, creative ideas, or opportunities to contribute to your vision. Feel free to reach out!

manojkumar.du.or.21@gmail.com

New Delhi, India

LinkedIn Profile