Neeraj kumar Taraka

Machine Learning Engineer

Hyderabad, Telangana, India10 yrs 9 mos experience

Most Likely To SwitchAI ML Practitioner

Key Highlights

Expert in NLP with extensive experience in AI/ML engineering.
Proficient in deploying scalable models using DevOps practices.
Innovative solutions for e-commerce and energy sectors.

Stackforce AI infers this person is a Machine Learning Engineer with expertise in NLP and DevOps for SaaS and E-commerce.

Contact

Skills

Core Skills

Natural Language Processing (nlp)Machine LearningBig Data

Other Skills

KubernetesNamed Entity Recognition (NER)Dependency ParsingOpenIEDeep LearningScalaMapReduceHBaseKafkaScaldingRocksDBPythonLarge Language Models (LLM)Generative AIPyTorch

About

With over 9 years of experience across data science, AI/ML engineering, and software development, I have honed my expertise in creating innovative solutions, particularly within the NLP domain. My work has involved complex projects like HSN-code classification, ranking in search engine, recommendation systems, and text understanding through relation extraction. I'm proficient in fine-tuning LLMs using tools such as MosaicML, PyTorch, and LightningAI. And always aiming to deliver cost-effective, high-performance models. In addition to my AI/ML skills, I bring a strong foundation in DevOps, particularly in automating and optimising workflows through Terraform, AWS, and CI/CD pipelines. This combination of AI/ML and DevOps expertise allows me to contribute to both the development and deployment of scalable, reliable solutions. Always Eager to take on new challenges and learning new things (any field).

Experience

10 yrs 9 mos

Total Experience

3 yrs 7 mos

Average Tenure

7 yrs 3 mos

Current Experience

Avalara

2 roles

Sr. Machine Learning Engineer

Apr 2024 – Present · 2 yrs 2 mos · Hyderabad, Telangana, India

Senior Software Engineer, AI/ML Engineer

Feb 2019 – Mar 2024 · 5 yrs 1 mo · Hyderabad, Telangana, India

I worked on extracting tax rate information from news articles and documents published by US states using NLP techniques like Named Entity Recognition (NER), Dependency Parsing, and OpenIE. Later improved the accuracy by fine-tuning large language models (LLMs).
I also developed tools using PDFBox and Camelot to enhance the detection and extraction of tables from PDF documents, making the process more efficient.
For e-commerce, I built deep learning models to automatically assign HSN codes to products based on their title and description. I created a unique approach that allows these models to work with very little training data and still make accurate predictions, even for codes with limited samples. I also developed image-based models as a backup when detailed product information is missing.
Additionally, I worked on improving search engine rankings for e-commerce products. This involved creating models for intent recognition and developing methods to provide confidence scores for search results, ensuring more relevant outcomes for users.
I ensured that all the models I built were smoothly integrated into production systems using Docker, Kubernetes, and shell scripts. I followed best practices in CI/CD and set up the necessary infrastructure with Terraform to make the deployment process efficient and scalable.

Natural Language Processing (NLP)KubernetesMachine Learning

Indix

Machine Learning Engineer

Dec 2017 – Feb 2019 · 1 yr 2 mos · Hyderabad, Telangana, India · On-site

I worked with technologies like HBase, Kafka, Scalding, and RocksDB to build pipelines for seeding, crawling, and parsing websites.
I built a batch-based crawl scheduling infrastructure that balances the freshness of data with the discovery of new products and items on e-commerce websites.
I replaced the legacy job with a dynamic scheduler using Kinetic Heaps, which improved the efficiency of the crawling process.
Additionally, I developed a metric tracking system for crawling that provides APIs to deliver time-bound aggregated results, making it easier to monitor and analyse the crawling performance.

ScalaMapReduceBig Data

Robert bosch engineering and buisness solutions

Data Scientist

Jul 2015 – Nov 2017 · 2 yrs 4 mos · Bengaluru, Karnataka, India

Developed algorithms for Non-Intrusive Appliance Load Monitoring (NILM, smart electric meters) problem to provide appliance specific energy consumption at home/commercial buildings by observing the aggregated power signal (main meter).
Worked on scrape reduction algorithms by analysing the data from production line in manufacturing industries.

PythonMachine Learning

Tata consultancy services

Software Engineer Internship

May 2013 – Jul 2013 · 2 mos · Pune Area, India

Designed and developed node-webkit application to migrate simple websites onto Adobe CQ5 platform
Built Chrome Extension to collect the html dom elements of a webpage and make cq5 readable template
Integrated Plugin and Nodejs modules to develop single executable software that scrapes complete website
Technologies Worked On: Javascript, NodeJs, Jquery, HTML, CSS