Developed detailed annotations for over 500 data points used in training AI models, identifying key areas for optimization that reduced output inaccuracies by 30%. Streamlined the AI-generated content review process through structured feedback loops, improving response clarity and consistency across automated interactions. Collaborated with cross-functional teams to ensure model outcomes aligned with business requirements, contributing actionable data insights for the business team.
AI Trainer
Summary
Developed detailed annotations for over 500 data points used in training AI models, identifying key areas for optimization that reduced output inaccuracies by 30%. Streamlined the AI-generated content review process through structured feedback loops, improving response clarity and consistency across automated interactions. Collaborated with cross-functional teams to ensure model outcomes aligned with business requirements, contributing actionable data insights for the business team.
Highlights
Developed detailed annotations for over 500 data points used in training AI models, identifying key areas for optimization that reduced output inaccuracies by 30%.
Streamlined the AI-generated content review process through structured feedback loops, improving response clarity and consistency across automated interactions.
Collaborated with cross-functional teams to ensure model outcomes aligned with business requirements, contributing actionable data insights for the business team.
Bachelor of Technology (B. Tech)
Electrical Engineering
→
Class-12 ICSE
Grade: 93.75 %
→
Class-10 ICSE
Grade: 95.8 %
achieving significant performance improvements in an Faster-RCNN, trained on 1,800+ HD images to detect automobile injuries (68%). Project Link
crafted effective strategies that directly improved risk prediction accuracy by enhancing existing models.
C, C++, Python, Swift, SQL, Java(basics).
GCP, AWS, NLP, Computer Vision, Sagemaker, vector.ai, Pytorch, LangChain, Hugging Face, Matlab, Postgres, Tableau, streamlit, MongoDB, GIT.
Time Series Analysis, OOPS, GPT, RAG, Terraform, Prompt Engineering, bash scripting.
Summary
Developed and implemented a robust sentiment analysis model using BERT and NLTK, processing data from over 60,000 tweets to provide actionable insights that influenced trading strategies for 25 high-profile stocks. Integrated advanced data-driven sentiment signals into portfolio risk analytics, enhancing the predictive accuracy and enabling more informed investment decisions across various asset classes.
Summary
Developed OpenELM-3B model for STEM/math tasks, achieving 2.36% higher accuracy than OLMo with 2× fewer pre-training tokens. Strong on ARC-c(42.24), MMLU(26.76), Hellaswag (73.28) benchmarks. Enhanced base model performance through targeted fine-tuning using LoRA, synthetic data and PRMs with Metal optimization and quantization, deploying final model via FastAPI for inference.
Summary
Implemented a multi-agent workflow to review insurance claims by analysing policy documents, damaged photos, and repair estimates, cutting manual review time by 40%. Trained the system using 500+ past claims data to improve accuracy. Integrated it with existing claims databases, allowing it to process 90% of routine cases without human input.