Lead DataScientist

B-Tech, JNTU Masters, BITS Pilani AWS GCP Tensorflow AI/ML Data Analysis LLM NLP Timeseries Forecasting MLOps

Experience

LEAD DATASCIENTIST Organization: Veltris Oct 2020 – Mar 2024 Handled multiple Predictive and Generative AI projects to deliver innovative solutions using Cloud services for multiple clients to improve their search mechanism and content management systems.

● Established a team from the ground up to develop minimum viable products (MVPs) centered around Generation AI applications, leveraging language models such as Mistral-13B and GPT-5 to cater to prospective clients.

● Managed a team of data scientists, providing guidance on project priorities, methodology, and best practices by implementing model and data governance frameworks.

● Implemented an intelligent search mechanism by providing tags using LLM’s based on the context provided and RAG approach to retrieve the context based and condition based search using self and entity based retrievers.

● Worked extensively using open source models like GPT-5, Llama, Mistral 13B etc and fine tuned them with various prompting techniques like zero-shot,one-shot and few shot learning to improve the prompt responses.

● Devised multiple strategies to finetune the models using PEFT, LORA and quantization techniques to improve the accuracy of the responses generated by the models.

● Performed model validation using prompt engineering and improved the model accuracy to 87%

● Partnered with software and hardware product development teams to accelerate and optimize future products in AI and High Performance Compute domains.

● Managed the end-end development and deployment of LLM’s to production ensuring the scalability of models by containerizing them on cloud infrastructure.

● Worked closely with internal engineering teams, external partners, customers, and ISVs to design and implement optimized, secure and responsible AI solutions on Cloud Platform.

● Reduced the model training time by 40%, by using the distributed training mechanism and containerizing the model components on cloud infrastructure using the AWS Kubernetes service.

● Reduced the high compute instances downtime by 10% and memory issues by monitoring model performance and GPU utilization metrics using Grafana dashboards.

● Presented findings to senior management in a clear and concise manner, facilitating data-driven decision-making.

Tools Used : Python, LLM, AWS Sagemaker, Tensorflow,Docker, Kubernetes, ML Flow, RAPIDS HPO, Distributed Training, Grafana

Sr. DATASCIENTIST Organization: Soctronics Jan 2020 – Sep 2020 Client: Genentech

● Led the development and implementation of Predictive AI solutions using cell culturing data to optimize drug discovery processes,by predicting the cell viability. Swathi Tatavarthy

+1-206-***-**** ******************@*****.***

Seattle,WA

Lead DataScientist with 10 years of experience in collaborating with cross-functional teams and ensuring the accuracy and integrity of data with actionable insights. Well-experienced in building AI products using LLM’s, NLP, Timeseries Forecasting and Predictive Analytics on Cloud Platforms.Skilled in machine learning, deep learning,statistics, problem-solving, and experience working in an agile development environment.

● Spearheaded the analysis of large-scale bioreactor and raman spectroscopy datasets, uncovering key patterns and insights using data transformation techniques like Modpoly, IModpoly, Log transformations etc

● Performed extensive EDA and feature engineering on spectrum data, offline and online variables like O2,CO2 etc resulting in a 70% improvement in model efficiency.

● Managed the end-end development and deployment of complex state-of-art and time series forecasting models to production ensuring compliance with industry standards and regulations for cell culturing.

● Improved the model accuracy by 10% using a meta-learning approach on the pre-trained CNN-LSTM model and experimented with multiple activations and optimizers specified from the MDPI research papers.

● Improved the model accuracy from 60% to 88% and reduced the RMSE to 0.12 using Hyper-parameter tuning frameworks like Keras-tuner and Optuna.

● Reduced the model training time by 40%, by using the distributed training mechanism and containerizing the model components on cloud infrastructure using the AWS kubernetes service. Tools Used : AWS,Python,Numpy,Scipy, Tensorflow,Kera-tuner,Bayesian&Gaussian Mixture Models,Kafka Streaming, TenorRT (Inferencing), TimeseriesModelling, CNN-LSTM, Grafana, Docker, Kubernetes, MLFlow SENIOR SOFTWARE ENGINEER Organization: Biarca Jul 2018– Jan 2020 Clients : Google Cloud, Admed

Worked on a flagship g-suite AI product ‘Biarca Patrol- a Cloud Security and Compliance Assessment Framework’ for Projects running on Google Cloud Platform and Worked on multiple AI projects using NLP, Semantic Search for ed-tech and staffing firms

● Design and develop AI products using Semantic search, Google Cloud Services and Infrastructure.

● Developed and implemented AI based document retrieval system using file metadata and tag generation. thereby improving the search efficiency of the on-prem file storage by 60%

● Implemented ML pipeline by stacking the outcomes of various classification models like (decision trees, random forest,xgbm).

● Improved the model accuracy from 60% to 88% and reduced the RMSE to 0.12 using Hyper-parameter tuning frameworks like Keras-tuner and Optuna on the transformer models.

● Implemented a multi-distributed training approach using Tensorflow framework for improving the model hyper- parameter tuning performance by 30%

● Developed and implemented threat models to measure the health of the Projects on GCP using Predictive AI and conducting regular security assessments and audits to validate the model's accuracy.

● Implemented vulnerability detection to the product by incorporating rules and intelligent notification mechanisms using recommender systems, thereby reducing the risk occurrence by 45%.

● Meet regularly with key customer-facing teams to collect and analyze user feedback to shape new ideas and user requirements.

● Conducted code reviews and implemented secure coding practices to mitigate potential security risks . Tools Used : Python, NLP, Transformers, BERT,RoBERT Models,Rest API, Docker, Kubernetes, ML Flow, GCP Storage buckets, Spacy, NoSql(Google Bigtable)

SOFTWARE ENGINEER Organization: OSI Technologies Jul 2015– Jun 2018 Client : Macys

Worked on implementing the e-commerce site of Macy’s and Bloomingde providing AI based recommendations and products based on the user preferences and browse history.

● Analyzed the user preferences and usage pattern to identify the important features from the historical data

● Developed clustering models to accurately provide recommended products, Ad’s for right set of customers

● Worked extensively on Statistical Modelling,EDA and collaborated with cross-functional teams to plan, design, develop, and procedures for building ML pipelines.

● Implemented stacked models and collaborative filtering to improve the accuracy of the models to 86%.

● Validated the model prediction capability with multiple user groups and refined the model accuracy using user rating & feedback mechanism.

Tools Used : Python,NLP, Spark ML, Clustering, Collaborative Filtering, Tableau Page 2

Contact this candidate