shashvat

Developed an Agentic AI bot for video processing—automating scene segmentation, summarization, and structured extraction via pytesseract, LLMs, and multimodal embeddings. Fine-tuned LLMs using Hugging Face Transformers and TensorFlow; built LangChain/LangGraph agents for dynamic, multi-step reasoning. Engineered scalable NLP pipelines on AWS SageMaker with Lambda for real-time inference by leveraging prompt engineering. Benchmarked and integrated APIs from Grok, LLaMA, OpenAI, etc for real-world NLP and multimodal tasks, while leveraging Azure services—Kubernetes and Functions—to optimize AI workloads for cost-efficiency and performance.

Engineered a secure authentication flow with social logins and email verification, achieving a 100% increase in user onboarding efficiency. Developed a meeting bot for Zoom and Google Meet, utilizing Zoom APIs, webhooks, SDK OAuth, and Google Meet APIs, syncing 100% of meeting data (video/audio) to AWS S3. Deployed a high-performance backend on AWS Elastic Beanstalk, scaling to support 10K+ concurrent users with a ReactJS frontend and a reliable PostgreSQL database.

Optimized application efficiency by implementing Firebase authentication for secure user access. Deployed a cryptographically secured payment gateway using Cashfree, enabling split payments, refunds, and seamless transactions. Spearheaded transition from LAMP to MEAN stack, reducing latency by 40% and improving database administration efficiency by 20%, while integrating DialogFlow chat bot and Google Analytics via Segment.