Ajayeswar Reddy
- +1(341)7328139
- ajayeswarb2a@gmail.com
- github
- San Francisco, USA

I'm a dedicated data science master's student at the University of San Francisco, and I'm currently an intern in AWS ML Labs' research division. With nearly 3 years of professional experience as a data scientist, I have a strong interest in designing, building, and deploying end-to-end scalable machine learning products. I have experience with a variety of tools and technologies and have a proven track record of collaborating with teams to deliver impactful results.
Work Experience
AWS ML Labs
Machine Learning Engineer
- Developing new algorithms and training pipelines to speed up neural network training. This research internship is with the USF data institute in association with the AWS ML labs.
GEP Worldwide Inc
Data Scientist
Work involved handling end-to-end ML and engineering tasks for the product OCR for invoices.
- Designed and implemented a sophisticated rule-based NLP pipeline to accurately extract critical attributes from invoices, achieving up to 100% accuracy for select attributes and an overall accuracy range of 70-100%.
- Revamped the architecture of the system by migrating the API from flask to FastAPI and transitioned from synchronous to asynchronous processing, which resulted in a significant improvement in the reliability and cut down the error rate for processing complex invoices by more than 90%
- Successfully decoupled a monolithic API into a microservices architecture, resulting in seven independent services with improved scalability, fault tolerance, and agility. Implemented a comprehensive testing plan and collaborated with the front-end team to seamlessly integrate new APIs with the existing user interface.
- Engineered the product using Docker and Azure Kubernetes serving ~10000 requests per day.
Data Science intern
Reverse Image Search
- Reverse Image SearchUsed pre-trained Res-net 50 architecture to encode the images and used modified K-nearest neighbor to classify the images and recommend the appropriate item.
- Reduced false-positive rate from 25% to 5% using classification models and utilized YOLO for accurate object detection, to identify multiple objects from one image and recommend accordingly.
Projects
Job recommendation – Developed an end-to-end distributed deep learning pipeline to match jobs with resumes. Link
- Scraped and stored job postings from multiple sources to GCP (Google cloud), preprocessed data using PySpark on Databricks, and stored collections in MongoDB.
- Generated embeddings for job postings and resumes using Word2Vec and Sentence Transformer models (large and small).
- Performed vector similarity search between job and resume embeddings using Pinecone achieving perfect accuracy.
ML from scratch – Implemented Decision Trees, Random Forest, Gradient Boosting, Adagrad, k-means clustering algorithms from scratch as part of coursework.
Recommendation systems – Developed Matrix Factorization recommendation system using PyTorch to predict movie ratings.
Search Engine – Implemented using Object Oriented hash table to retrieve matched documents from a corpus in constant time.
Mixed Code Sarcasm Detection – Implemented using Object Oriented hash table to retrieve matched documents from a corpus in constant time. Link
Certifications
-
Deep Learning SpecializationCourses: Neural Networks and Deep Learning, Hyperparamater tuning and optimization, Structuring Machine Learning projects, Convlutional Neural Networks, Sequence ModelsCoursera - 2020
-
TensorFlow in practice specializationCourses: Introduction to TensorFlow, Convlutional Neural Networks in TensorFlow, Natural Language Processing in TensorFlow, Sequences, Time Series and PredictionCoursera - 2020
-
python for everybody SpecializationCourses: Python Data Structures, Using Python to access Web Data, Using Databases with PythonCoursera - 2020
-
Machine Learning: Algorithms in the Real World SpecializationCoursera - 2020