At Innodata, I evaluate AI-generated responses for accuracy, reasoning, relevance, and instruction-following. I also support AI quality workflows through prompt refinement and multimodal evaluation of English and Hindi text, images, audio, and video.
At Ethara AI, I evaluated 1,000+ LLM responses for quality, reasoning, and correctness, and refined and standardized outputs to improve response consistency across tasks.
At Celebal Technologies, I built CI/CD pipelines with GitHub Actions and Docker, then deployed containers to AWS EC2/ECR. This reduced deployment time by about 30%, and I automated multi-stage workflows and secret handling.
At Suvidha Foundation, I created a labeled headline dataset and used XGBoost and logistic regression models to improve categorization accuracy by 20%. My projects include a multilingual BERT-based resume screener, a LangChain and ChromaDB RAG API, and a latent-diffusion steganography system.
