Aamir Ahmad Ansari

Versatile Data Scientist specializing in Deep Learning, Machine Learning, Reinforcement Learning, and MLOps. Skilled in statistical and algebraic analysis, with expertise in developing robust models and end-to-end machine learning pipelines. Highly organized and self-motivated, delivering innovative solutions to complex data challenges. Committed to advancing AI technologies and driving meaningful insights in the rapidly evolving field of data science.

• Date of Birth: 06/11/1998
• Nationality: Indian
• Marital Status: Unmarried

2023-12 - 2024-01, Dear Diary, New Delhi, India

• An implementation of a chat bot using Large Language Models and Streamlit which can act as your own personal digital diary.
• Integrated Google DeepMind's Gemini and Microsoft's Phi-2 models.
• Added SqLite Database for chat retention.
• Custom post-processing for Phi-2 outputs.
• User-friendly interface built with Streamlit.

2022-08 - 2022-12, Emotion Detection and Analysis, New Delhi, India

• It is a web application which helps digital public speakers to analyze their audience's behavior. It detects facial emotions, performs sentiment analysis on the speaker's speech and generates a complete report.
• The report contains time series analysis as intuitive graphs that shows the speakers what went right or wrong. We use SOTA models like YOLO-V7, Sentence BERT and OpenAI Whisper.
• Presentation on YouTube, starts at 16:15 mins. Youtube:https://www.youtube.com/watch?v=G5-u3YE2TSQ&t=931s | Github: https://github.com/aamir09/FinalSubmissionAI5

Masters AI & ML, Univ.AI

2023-10 - 2024-05, Shape Morphing Evolution in Higher-Dimensional Space, New Delhi, India, (Remote, Part Time)

• Volunteering in Professor Ivo Sbalzarini's research group to track and cluster trajectories of various shapes that evolve over time.
• Spearheaded the development of cutting-edge data modeling techniques, consistently identifying and implementing optimal solutions for the given time series dataset using the following techniques: Elliptical Fourier Transform, Physics Informed Neural Networks, Neural ODE's, LSTMs.
• Conducted in-depth data analysis, extracting valuable insights using latent space analysis of the AutoEncoders fit on the data .
• Delivered weekly presentations featuring compelling data visualizations, effectively communicating findings and recommendations to supervisors.

2023-05, Publication, Tree based Convolutional Neural Networks for Image Classification, ICIDSSD'22