Oct 2022 – Present · 3 yrs 7 mos

• 1. Tailored Llama-13b-chat for distinct tasks, refining performance and output format against custom datasets.
• 2. Spearheaded the development of a semantic search application leveraging proprietary LLM and embedding model.
• 3. Enhanced and optimized entity-disambiguation pipeline by integrating PySpark, RabbitMQ, and ScyllaDB.
• 4. Implemented scalability measures and executed disambiguation pipeline over 100sM data records within a few days.
• 5. Engineered a solution to extract information structurally from digital congress documents, significantly reducing processing time by two weeks ultimately saving recurring costs.
• 6. Orchestrated the deployment of Federated Learning system using NVFlare on in-house biomarker prediction algorithm, achieving comparable accuracy levels, opening our way for processing private data within compliance.
• 7. Optimized translation pipeline utilizing multiple GPUs and FashText, slashing processing time from days to hours.
• 8. Experimented and developed synonym classification for biomedical entities using LLMs.
• 9. Ranked emerged and identified emerging Key Opinion Leaders using a diverse set of 30+ carefully crafted features.

Mar 2019 – Sep 2022 · 3 yrs 6 mos

• 1. Pioneered a methodology to seamlessly merge multiple ontologies from over 10 sources, encompassing 2.5M entities, while safeguarding against cycles or duplicates and preserving hierarchy relationships and synonyms.
• 2. Enhanced a knowledge graph of 3.4B nodes and 1.7B connections on PyTorch-BigGraph to generate graph embedding and RandomForestClassifier to build a classification model for entity disambiguation.
• 3. Architected a system capable of extracting biological entities from literature and establishing contextual relationships, culminating in the creation of a comprehensive biological knowledge graph.
• 4. Developed an algorithm to autonomously generate an ontology for category-related concepts leveraging multiple databases and transformer-based techniques.
• 5. Crafted a custom pipeline for transliterating biomedical entities, and names in four different languages using LMs.

Aug 2016 – Feb 2019 · 2 yrs 6 mos

• 1. Achieved reduction in manual effort by implementing automation for extracting entities from targeted web pages using ML.
• 2. Developed extraction pipeline utilizing CV techniques to extract variables and tables from PDFs.
• 3. Improve performance for the data transformation processing module in the pipelines.
• 4. Created in-house API services and Python clients for geolocation normalization, abbreviation expansion, and affiliation extraction. Ultimately build affiliation normalization algorithm.

May 2014 – Jul 2014 · 2 mos · New Delhi, Delhi, India

• 1. Developed web forms using C# along with SQL database integration for storing data.
• 2. Formulated a strategy and submitted a report for SEO of the company's website by analysing factors like HTML-tag, website format, meta-tag optimization, trends analysis of competitor websites, etc.

Jul 2011 – Jul 2016 · 5 yrs · Greater Delhi Area

• Completed Integrated MTech degree in Mathematics and Computing.
• The courses in this degree were majorly focused around the mathematics involved in any solution/hypothesis. In elective course choices I enrolled for Computer Science courses.