Oct 2023 – May 2025 · 1 yr 7 mos · Gurugram, Haryana, India · On-site

• High-Speed Embedding Deployment: Migrated the embedding model from an NVIDIA-based instance to an AWS Inf1 instance and deployed an API with Nginx reverse proxy, boosting inference speed by 8 times and enabling 10M+ data embeddings per day.
• Prompt Management: Developed a centralized LLM prompt management module for product-wide applications (Tagging Accuracy, Key Highlights), enabling efficient testing and tracking of F1 and accuracy scores for each prompt.
• Tagging Accuracy & Key Highlights: Developed a high performance tag correction and key highlight extraction pipeline using LLMs to verify in-house model tags, boosting F1 score to 95%+ and ensuring precise key point extraction.
• Contify Copilot: Developed a RAG pipeline that seamlessly answered user queries using pre-scraped, processed data and user uploaded attachments, enhancing the user experience with fast, relevant responses.
• Language Translation: Deployed a translation model to convert non-English content to English, enabling advanced NER and business tagging with machine learning—boosted non-English content crawling by 36x and eliminated costly external translation services

Apr 2022 – Sep 2023 · 1 yr 5 mos · Gurugram, Haryana, India · On-site

• NER Tagging: Upgraded the NER tagging pipeline from NLTK to a state-of-the-art BERT model, boosting accuracy by 3 times and performance by 4 times.
• Tailored News Curation: Deployed client-specific BERT-based multiclass classification models to filter business and non-business news, ensuring personalized and relevant content delivery for each client.
• Topic and Industry Classification: Developed two high-performing BERT-based multilabel classification models achieving over 90% F1 score, deployed for classifying content into industry-specific and general topics categories, handling more than 1 million daily requests.
• Cost-Efficient ML Deployment on AWS: Engineered the conversion and deployment of ML models into torch-compatible formats for AWS Inf1 instance that reduced company costs for AWS EC2 instance usage by up to 80%.
• Contify-MS Teams Bridge: Integrated the Contify platform with MS Teams, enabling seamless update sharing to teams and channels, enhancing collaboration and communication efficiency.