Mayank Bhagat
Creative AI Engineer
I build intelligent
digital systems.
Bridging AI research
With
real-world impact.
ABOUT
ME.
I am a Creative AI Engineer based in India, specializing in building intelligent systems that bridge the gap between complex algorithms and intuitive human experiences.
With a deep focus on Generative AI, Computer Vision, and Full-Stack Development, I craft digital solutions that are not just functional, but transformative.
When I'm not coding, I'm exploring the frontiers of computational creativityor contributing to the open-source community.
Experience
Machine Learning Intern
- Led a team of 5 to develop CliniScan, an AI-based lung abnormality detection system for radiologists.
- Built a multi-label chest X-ray classification pipeline using EfficientNet-B0 CNN on 18,000+ VinBigData DICOM images.
- Achieved strong clinical performance with Macro F1 = 0.85, Micro F1 = 0.89, and Macro AUC-ROC = 0.92.
- Implemented lesion localization using YOLO (Ultralytics) achieving mAP@0.5 = 0.84.
- Integrated Grad-CAM visualization and deployed a real-time inference pipeline.
Summer Industrial Internship – Generative AI
- Implemented enterprise-grade Generative AI pipelines using IBM Watsonx.ai, improving reliability by 27%.
- Compiled and optimized RAG workflows integrating LLMs, increasing retrieval precision by 18%.
- Designed custom prompts and optimized context windows, reducing token usage by 22%.
- Developed MediBot, an AI-powered medical triage chatbot using RAG, achieving 92%+ routing accuracy.
- Enhanced medical response safety by integrating retrieval-backed knowledge base, improving factual consistency by 35%.
Technical Arsenal
Machine Learning & AI
Frameworks & Architectures
Languages
Libraries & Tools
Cloud Services
NLP & Speech
Recognition
Publications
PIPO: AI-Based Voice Assistant
AI Global Conference 2025 (Sponsored by IEEE)
Presents a lightweight, Python-based desktop voice assistant integrating Whisper STT, mBERT for intent classification, SpaCy for NLP processing, and GPT-based response generation. The system supports multimodal input (voice and text), real-time command execution, and desktop automation with high intent-classification accuracy (94%). Experimental results demonstrate improved efficiency, conversational flow, and performance compared to existing Python-based voice assistants.
M. R. Bhagat, H. Kalwar, D. S. Jamwal, S. Purohit, R. Singh and T. Maskara, "PIPO: AI-Based Voice Assistant," 2025 2nd Global AI Summit - International Conference on Artificial Intelligence and Emerging Technology (AI Summit), Noida, India, 2025, pp. 969-974, doi: 10.1109/AISummit66170.2025.11411619. keywords: {Weather;Personal voice assistants;Smart homes;Software;Internet;Text to speech;Reliability;Artificial intelligence;Python;Videos;"Artificial Intelligence";"Desktop Assistant";"Python";"Text-to-Speech";"Virtual Assistant"}
Honors & Awards
SIH Finalist
Smart India Hackathon 2025
Spearheaded a team of 6 and Ranked in top 45 teams among 950+ teams. Problem Statement: SIH25045 — CropSure (Agritech-based Yield Monitoring Solution)
IBM Cloud Certification
IBM Private Limited
Acquired a Cloud Computing Certificate from IBM Private Limited and Recognized for exceptional contribution to the Watsonx.ai generative AI deployment project.
View CertificateMicrosoft Azure Certification
Microsoft
Acquired a Cloud Computing Certificate from Microsoft DP-900 for Azure Fundamentals.
View CertificateApplied Machine Learning using Python
Michigan University
Received a certificate for completing the Applied Machine Learning using Python course with score of 98%.
View CertificateAWS Certified Cloud Practitioner
Amazon Web Services
Earned the AWS Certified Cloud Practitioner certification.
View CertificateOriginals
Selected works and experiments.
AI Lawyer
Engineer a legal assistant using Generative AI concepts. Achieved 85%+ clause retrieval accuracy using dense vector search. Orchestrated a memory-augmented pipeline and RAG workflows incorporating LangChain and Hugging Face LLMs, reducing hallucination rate in context-aware legal conversations by 20% and improving user satisfaction.
Overview
Engineer a legal assistant using Generative AI concepts. Achieved 85%+ clause retrieval accuracy using dense vector search. Orchestrated a memory-augmented pipeline and RAG workflows incorporating LangChain and Hugging Face LLMs, reducing hallucination rate in context-aware legal conversations by 20% and improving user satisfaction.
Rojgar Boat
The website is a React.js-based frontend that provides users with an interactive interface for job search and discovery. It integrates an AI search engine to enhance query relevance and result understanding, while MongoDB serves as the backend database. Features full text search and filtering.
Overview
The website is a React.js-based frontend that provides users with an interactive interface for job search and discovery. It integrates an AI search engine to enhance query relevance and result understanding, while MongoDB serves as the backend database. Features full text search and filtering.
Pipo : AI based voice assistant
Developed multilingual voice assistant for real-time automation tasks with implementation of ASR and TTS systems achieving 1.2s response latency.Integrated commands for system-level achieving overall 92% accuracy with multimodals( BERT, SpaCy, GPT-2 ).
Overview
Developed multilingual voice assistant for real-time automation tasks with implementation of ASR and TTS systems achieving 1.2s response latency.Integrated commands for system-level achieving overall 92% accuracy with multimodals( BERT, SpaCy, GPT-2 ).
Hospitality and Health Hub
Composed Java Swing-based HMS with MySQL backend supporting CRUD, appointment scheduling, and patient data management. Streamlined database structure for faster multi-user access and reduced query latency.
Overview
Composed Java Swing-based HMS with MySQL backend supporting CRUD, appointment scheduling, and patient data management. Streamlined database structure for faster multi-user access and reduced query latency.
Smart ops
An agentic brain for businesses based on agentic AI. Built for businesses to manage their workloads by logistic services (find the perfect cost-efficient route by all transport ways while saving carbon emissions and cost) and finest AI inventory management that suggests products to add according to latest ups and downs in the market. Features an analysis window for overall business review with multi-agents working on it.
Overview
An agentic brain for businesses based on agentic AI. Built for businesses to manage their workloads by logistic services (find the perfect cost-efficient route by all transport ways while saving carbon emissions and cost) and finest AI inventory management that suggests products to add according to latest ups and downs in the market. Features an analysis window for overall business review with multi-agents working on it.
QuantHack Core
An advanced business workload manager that leverages quantum computing to produce lightning-fast and mathematically perfect outputs. By utilizing cutting-edge quantum libraries instead of traditional Agentic AI, it provides unparalleled processing speed and an ultra-secure environment for logistics optimization and data analysis.
Overview
An advanced business workload manager that leverages quantum computing to produce lightning-fast and mathematically perfect outputs. By utilizing cutting-edge quantum libraries instead of traditional Agentic AI, it provides unparalleled processing speed and an ultra-secure environment for logistics optimization and data analysis.
© 2026 Mayank Bhagat