Sep 2023 – Feb 2026 · 2 yrs 5 mos · Gurugram, Haryana, India · Remote

• Designed a serverless text-to-SQL solution using AWS Bedrock, Glue, Athena, and Lambda, enabling non-technical users to query S3 CSV data in natural language, significantly reducing dependency on SQL and BI teams.
• Built an AI-driven multi-agent contract review system using LangGraph and GPT to automate the verification of insurance policy text against mainframe data. Deployed via Docker on AWS, the project achieved a verifiable 90% reduction in manual CSR review effort.
• Fine-tuned Meta Llama‑3 on AWS SageMaker with PEFT/LoRA for insurance domain adaptation using policy documents, claims, and FAQs, improving relevance and consistency of domain-specific generations.
• Architected Row-Level Security (RLS) in Weaviate using an MCP Server, enabling manager–subordinate inheritance and relationship-based access filtering. Designed a secure retrieval layer with server-side RLS enforcement to ensure strict data isolation and RBAC compliance.
• Developed a domain-aware English–French translation pipeline leveraging RAG for specialized terminology, improving translation accuracy and reducing post-editing effort for business teams.

Mar 2021 – Sep 2023 · 2 yrs 6 mos · Gurugram, Haryana, India · Hybrid

• Developed RAG-based banking customer support chatbot using LangChain, OpenAI GPT-3.5, and FAISS to retrieve answers from internal FAQs and policy manuals, improving query resolution accuracy and reducing average response time.
• Built Flask-based AI compliance assistant with FAISS embeddings and GPT-3.5 to support regulatory Q&A with cited sources, reducing manual compliance research effort and associated operational costs.
• Engineered an automated query classification system using BERT to categorise multi-channel banking queries (credit cards, loans, complaints), significantly reducing manual triage time for operations teams.

Mar 2020 – Feb 2021 · 11 mos · India

• Engineered a hybrid talent recommendation model using NLP skill embeddings & collaborative filtering with explainable ranking to match employees to internal roles.
• Built a training recommendation engine that matched employees to training programs based on skill assessments, career goals, and course metadata, improving relevance of suggested upskilling paths. boosting employee skill development and overall learning engagement.
• Developed an AI-driven skill extraction system to accurately identify and classify skills from resume, managing end-to-end text annotation for NER. Built a BERT-based NER model in PyTorch and designed the full job-parsing workflow using Deep Learning and NLP techniques. Deployed the complete solution on AWS Cloud.

Nov 2017 – Mar 2020 · 2 yrs 4 mos · India · On-site

• Built a regression-based health insurance pricing model to predict member-level 12‑month claim cost using demographics, policy details, utilisation history, and health risk indicators, improving expected loss prediction accuracy and enabling more granular premium setting.
• Integrated sentiment analytics on a leading logistics company’s customer experience platform by developing a robust sentiment analysis solution that processes customer feedback from multiple channels—including surveys, social media, and support tickets—enhancing the organisation’s capability to accurately detect and respond to positive, negative, and neutral sentiments.
• Clean, aggregate, pre-process and organise structured and unstructured data.
• Analysed and manipulated the data for Exploratory Data Analysis (EDA), and present the data in a clear manner.

May 2014 – Oct 2017 · 3 yrs 5 mos · Gurugram, Haryana, India

• Worked on lead management system and mobile Business applications.
• Collaborated with cross-functional teams to define, design, and ship new features.
• Conceptualized, designed, developed and deployed applications for android based mobile phones.