AI Backend Engineer / GenAI RAG Engineer

Built a production-grade Python image AI virtual try-on and dress-swapping pipeline that transfers garments from source images onto target persons while improving realism, alignment, and identity preservation.

Developed FastAPI backend services for model inference, request validation, asynchronous task handling, structured logging, and client integration. Implemented OpenCV and NumPy preprocessing for human parsing, segmentation, garment masking, pose alignment, normalization, mask refinement, blending, and post-processing.

Deployed scalable AI services with AWS EC2, Lambda, S3, SageMaker, Docker, GCP, and cloud orchestration for model serving and asset management.

Developed a multi-agent educational AI platform using Python, FastAPI, Redis, and PostgreSQL for assessment generation, feedback analysis, lesson recommendations, and teacher reflection workflows.

Built event-driven orchestration with Redis Streams and implemented collaborative, stateful workflows using CrewAI and LangGraph. Exposed agent capabilities through FastAPI REST APIs for downstream product integration.

Integrated Google Classroom, Canvas, and Microsoft Teams through APIs and LTI to synchronize assignments, grades, submissions, and rosters. Used AWS Bedrock and Claude 3.5 Sonnet for reasoning, lesson generation, contextual assistance, and feedback workflows.

Developed Python desktop and backend AI applications for deepfake detection, computer vision inference, voice AI, and model evaluation. Created an offline deepfake detection application using OpenCV and reduced inference latency by 40% through preprocessing and model-execution optimization.

Trained and evaluated ViT and MC3-18 models, achieving 98% accuracy for violence detection with cross-dataset validation. Deployed deep learning workflows on Google Vertex AI for automated training, evaluation, experiment tracking, and versioned model pipelines.

Built the DeepfakeCheck.in Django portal, enabling more than 500 testers to upload and analyze deepfake images, videos, and audio. Also implemented voice cloning, synthetic speech, STT/TTS, and LLM-driven conversational voice systems.

Developed full-stack AI applications with emphasis on Python backend development and image-processing workflows.

Optimized FastAPI backend services and reduced model inference response time by 45% through API and processing improvements. Built AI-powered classification features and collaborated with cross-functional teams on backend integration and deployment workflows.

Built real-time machine learning applications in Python and deployed them using Streamlit and cloud services.

Designed reusable ML pipelines and lightweight API-driven workflows to support rapid experimentation, testing, and deployment.

CGPA: 8.3/10.