Sep 2022 – Nov 2025 · 3 yrs 2 mos · Gurugram, Haryana, India · On-site

• Developed a Generative AI-driven multi-lingual solution for automated review of healthcare promotional content. Utilized Adobe Extract API to parse documents and implemented a pipeline leveraging GPT-4o for detecting grammatical errors, misinformation such as exaggerated claims and downplayed safety risks, active identification of missing brand specific logos in the promotional documents. Employed advanced prompt engineering techniques to compare promotional materials against verified sources, enhancing compliance and reducing manual review efforts.
• Led a fast-paced project to identify high-risk patients likely to discontinue drug therapy, delivering results within 6 weeks. Built and deployed an XGBoost model with feature selection via Random Forest and SHAP analysis, achieving actionable insights that informed business interventions. Oversaw end-to-end development with MLflow tracking and conducted workshops to guide client-facing teams on data-driven strategies.
• Developed and deployed a Bayesian Poisson-Gamma model using MCMC techniques to forecast patient enrollment at the trial-site level, aiming to reduce clinical supply waste by 5% annually. Collaborated with MLOps teams to design and implement the deployment architecture, leading to an estimated $15.15 million in savings by 2024.

May 2022 – Aug 2022 · 3 mos · Noida, Uttar Pradesh, India

• Developed real-time text classification model to classify the queries raised on Bill of Exports during customs review process on foreign imports.

Jul 2019 – Apr 2022 · 2 yrs 9 mos · Noida, Uttar Pradesh, India

• Developed an unsupervised NLP pipeline to extract caller pain points from call transcripts. Utilized pre-trained BERT embeddings for semantic clustering based on cosine similarity, followed by TF-IDF analysis to generate significant n-grams. This approach provided actionable insights into distinct issues discussed during calls, enhancing customer support strategies.
• Designed an NLP-based model to classify chat inquiries from health insurance members, aiming to uncover key drivers behind member outreach. Used transformer models (e.g., BERT) to accurately categorize chats into intents such as benefits, claims, prior authorizations, billing, and prescriptions. The solution enabled active identification of opportunities in patient's healthcare journey and provided business teams with actionable insights to reduce repeat contacts and enhance member experience.
• Built an analytical solution to measure the digital portal’s impact on reducing member calls. Used K-Means clustering and Euclidean Distance Risk Matching with GBM-selected variables to segment members and identify matched control groups. Validated clustering accuracy and conducted call savings analysis to deliver actionable business insights.
• Developed a repeat caller prediction model to identify high-risk members driving multiple calls. Leveraged advanced feature engineering and machine learning techniques to improve accuracy. Delivered insights that enabled targeted interventions, reducing call volume and enhancing member experience.

Sep 2018 – Jun 2019 · 9 mos · Noida, Uttar Pradesh, India

• Analyzed the efficiency of a newly implemented chat service compared to traditional call support by examining follow-up rates and agent-customer interaction times. Identified that approximately 7% of total chat duration was spent by agents waiting for customer responses, even before customers exited the chat. This insight led to the adoption of an asynchronous chat model, enhancing agent productivity and customer experience.

Jun 2016 – Aug 2018 · 2 yrs 2 mos · Noida, Uttar Pradesh, India

May 2015 – Jun 2015 · 1 mo · Bhagalpur Area, India

• Joined BSNL as a trainee, to get a basic insight of the working of telecomm. companies. The training consisted of basic theory classes on the various technologies used in the different communication standards as well as exposure was given to the equipment used. Thus, the topics covered were GSM, CDMA, FDMA, TDMA BROADBAND, MOBILE COMMUNICATION and LANDLINE COMMUNICATION.