Headline

Senior Machine Learning Engineer

Location

Patna

AI stack

aistudio

antigravity

claude

github copilot

hugging face

Work experience

Senior Software Engineer at Provaantech, Oct 2023 - Present · 2 yrs 10 mos, India, • Developed and optimized GCP vertex AI pipelines, achieving ~ 90% reduction in training costs. • Hands-on expertise in generative AL, LLM fine tuning & langchan based workflow automation. • LLM_pdf_ask, LLM_summarization, LLM_chat endpoint created to use LLM models in real time. • Rich Experience in managing entire data science project life cycle and involved in all phases, including data extraction, data cleaning, statistical modeling and data visualization, with large datasets of structured and unstructured data. • Hands - on experience in Generative AI LLM and Machine Learning algorithms such as GPT 3.5, Linear Regression, GLM, CART, SVM, KNN, LDA/QDA, Naive Bayes, Random Forest, SVM, Boosting, K-means Clustering, Hierarchical clustering, PCA, Feature Selection, Collaborative Filtering, Neural Networks and NLP. • Professional working experience with Python 3 libraries including MatplotLib, Numpy, Scipy, Pandas, Beautiful Soup, Seaborn, Scikit-learn, OpenCV and NLTK for analysis purpose. • Working experience in version control tools such as Git, SVN, Clearcase to maintain the source code. • Employing various SDLC methodologies such as Agile and SCRUM methodologies. • Working knowledge in RestFul API automation using Django & flask framework, &Postman tool. • Implemented new features and optimized existing code for the platform using Kubernetes & Docker.

Senior Associate at Deltacubes, Nov 2018 - Sep 2023 · 4 yrs 11 mos, India, Client: PMI healthcare, Bristol Myers Squibb (BMS), Merck, J&J, etc Domain: Pharmacovigilance, healthcare Project 1: LLM POC and Medical named entity recognition using LLM Responsibilities: ▪ Performed a poc on llm_chat_pdf, llm_Q_&_A, llm_summarization, llm_translate. ▪ Extract the named entities such as adverse event, suspect drug, dug dosage, medical history from narrative using openai llm model and langchain framework. ▪ GPT 3.5 used from development to deployment. ▪ Restrict the output format of response using RespsoneSchema. Project 2: MedDRA AIEncoder: Medically coding reported term. Responsibilities: ▪ Data collection from various customers and preprocessing to assess data bias and perform necessary data augmentation. ▪ Generate embeddings using Universal Sentence Encoder and AzureOpenAI Embeddings to generate sentiment embeddings in vector terms and stored them in an Elasticsearch database. ▪ Designed Elasticsearch node for retrieving relevant index data using Euclidean distance (KNN) and cosine-similarity. ▪ Developed a ML ranker model (Deep Learning BERT TensorFlow 2) using historical data to provide scores based on the similarity of reported terms. ▪ Integrated with Amazon Kafka (MSK) to receive live ground truth data from the pipeline for feedback mechanism and visualize the performance. ▪ Deployed end to end service using django framework.

Software Engineer Project Lead at IBM, Jan 2017 - Aug 2019 · 2 yrs 8 mos, India, Lines and Checkboxes detection from pdf/images. Responsibilities: ▪ Django Rest api endpoint is designed to extract pertinent information from unstructured data sources such as medical reports, medical appointment documents, and scanned documents. This is achieved through the utilization of Computer Vision (OpenCV) and Deep Learning techniques to convert the data into structured form. ▪ Keras Inception-v3 deep learning CNN model used as a based model, which was further refined by incorporating historical labeled images from multiple safety forms. ▪ Performance enhancements led to a significant reduction in checkbox detection time, from 25 seconds to 5 seconds. ▪ Implemented the lines endpoint to detect vertical and horizontal lines on safety forms, facilitating structured data extraction. ▪ form_detection endpoint to identify the category/format of medical forms.

Senior Software Engineer at GlobalLogic, Oct 2015 - Jan 2017 · 1 yr 4 mos, India, Project 1: Predict Track circuit modelling environment based on track input voltage and frequency. Project 2: Train speed recognition from speedometer snapshot. Project 3: Error margin optimization of track circuit border and train detection point. ▪ Extracting data from various databases by implementing ETL process, wrote and optimized SQL queries to perform data extraction. ▪ Used OpenCV module to do Image processing of 2 Million images in each iteration. ▪ Conducted analysis on assessing customer consuming behaviors and discover value of customers with RMF analysis; applied customer segmentation with clustering algorithms such as K-Means Clustering and Hierarchical Clustering. ▪ Used Python 3 (numpy, scipy, pandas, scikit-learn, seaborn, NLTK) and Spark to develop variety of models and algorithms for analytic purposes.

Senior Software Engineer at SPAN Infotech (India) Pvt. Ltd., May 2012 - Sep 2015 · 3 yrs 5 mos, India

Education

Visvesvaraya Technological University