Principal Software Engineer & Lead Data Scientist with 5.5+ years of experience in fast-paced startups and software engineering. Expert in building and deploying scalable Machine Learning, Deep Learning, and Generative AI solutions. Proven track record of solving complex problems and delivering impactful solutions through advanced technologies.

Currently leading AI/ML initiatives at University Living, previously Senior ML Engineer at Contify and ML Engineer at J&F. Specialized in LLMs, Multi-Agent Systems, RAG pipelines, and production-scale ML deployments.

• AI Voice Assistants - 50% improvement in conversion rates through human-like interactions
• Conversational AI Chatbots - 40% reduction in manual support workload
• Real-Time NLP Analytics - 100% automation in call analysis and insights extraction
• Multi-Agent Systems - F1 score improvement from 0.85 to 0.95 using self-correcting agents

🎙️ AI Voice Assistant for Accommodation Booking

• Developed AI-driven outbound voice assistant using Twilio, FastAPI, Elasticsearch, OpenAI LLMs
• Achieved 50% improvement in conversion rates through human-like conversational interactions
• Implemented dynamic lead scoring for enhanced customer targeting

🤖 AI Chatbot for Accommodation Booking

• Built conversational AI chatbot using Python, OpenAI, FastAPI, SQLite, Redis, Elasticsearch
• Reduced manual support workload by 40% with 24/7 intelligent user engagement
• Integrated multi-modal capabilities for enhanced user experience

📊 Real-Time NLP-Driven Call Intelligence Platform

• Engineered real-time analytics platform using NLP and FastAPI
• Achieved 100% automation in call analysis extracting intent, sentiment, and booking probability
• Built interactive dashboards powered by PostgreSQL for data-driven sales decisions

🏷️ Efficient Multi-Label Classification Models

• Developed high-performance models using distilled BERT and BERT architectures
• Achieved 90%+ F1 score for industry-specific and general topic classification
• Handled 2.5 million daily requests ensuring robust content categorization

🔍 Prompt Evaluation & Production Monitoring System

• Designed scalable evaluation framework assessing prompts across hundreds of thousands of data points
• Significantly reduced hallucination rates in production LLM outputs
• Implemented real-time monitoring and alerting for continuous accuracy tracking

🧠 Contify Copilot (RAG Chatbot)

• Built Retrieval-Augmented Generation (RAG) pipeline integrating pre-scraped and user-uploaded documents
• Enhanced user experience with fast, context-aware responses from historical and new datasets
• Optimized retrieval mechanisms for precise information extraction

🔄 High-Accuracy Deduplication Pipeline

• Led design and deployment achieving 98% accuracy in duplicate elimination
• Reduced data noise and improved integrity for downstream processing
• Implemented scalable architecture handling millions of documents

🦙 LLaMA-3.1 Finetuning & Deployment

• Finetuned LLaMA-3.1 8B model using QLoRA for text classification
• Deployed with vLLM for optimized inference speed and reduced memory usage
• Achieved 3x reduction in serving cost with low-latency predictions

🤝 Multi-Agent Tagging System

• Designed self-correcting multi-agent system using LLMs
• Agent 1 validates tags, Agent 2 resolves ambiguities with high confidence updates
• Boosted F1 score from 0.85 to 0.95 through intelligent collaboration

🕸️ Knowledge Graph-Powered Entity Extraction

• Developed NLP system extracting entities and relationships using LLMs
• Built Neo4j knowledge graphs from 100K daily documents
• Improved strategic decision-making accuracy by 40% through complex Cypher queries

🌐 In-house Translation Model

• Deployed proprietary translation model converting non-English content to English
• Increased non-English content crawling by 36x
• Eliminated costly third-party translation service dependencies

☁️ Cost-Effective AWS Deployment

• Successfully deployed ML models on AWS Inf1 instances
• Achieved 80% cost reduction while maintaining high performance and scalability

📱 Real-time Custom Image Classification

• Led automated ML model training system for iOS apps
• Utilized Turicreate, PyTorch, TensorFlow, Flask for optimization

🏗️ Real Estate BIM 7D Automation

• Streamlined floorplan detection, document classification, data extraction
• Implemented room calculations, symbol detection, iOS sensor identification
• Used Turicreate, Pytesseract, TensorFlow, OpenCV, Flask

🎯 Custom Object Detection Training

• Created user-friendly solution for customized object detection models
• Leveraged YOLO v3 algorithm and Flask for real-time detection

• 🌟 Rising Star @ Contify (2022) - Outstanding performance and innovation
• 👨‍🏫 Gurudronacharya Award @ Contify (2022) - Exceptional mentoring and leadership
• 📈 50% Conversion Rate Improvement - AI Voice Assistant implementation
• 💰 80% Cost Reduction - AWS Inf1 deployment optimization
• 📊 98% Accuracy - High-performance deduplication pipeline

• Machine Learning Certification - Coding Blocks
• Problem Solving Certification - HackerRank
• B.Tech Computer Science - Jamia Hamdard University (CGPA: 8.2)

• 2.5M daily requests handled by multi-label classification models
• 100K daily documents processed by knowledge graph pipeline
• 36x increase in non-English content processing
• 3x cost reduction in model serving infrastructure

• 90%+ F1 score in multi-label classification
• 98% accuracy in deduplication pipeline
• F1 score improvement 0.85 → 0.95 via multi-agent systems
• 40% improvement in strategic decision-making accuracy

• 50% improvement in conversion rates (Voice Assistant)
• 40% reduction in manual support workload (AI Chatbot)
• 100% automation in call analysis and insights
• 80% cost reduction in cloud infrastructure