AI Consultant

Jan'22 - Present

Clients (Europe & US)

AI Career Assistant LLM Agent with Semantic Kernel, RAG, and Plugin Architecture

• Designed and implemented a multi-step agent using Microsoft Semantic Kernel that orchestrates multiple AI plugins to extract, refine, and enrich user queries using Azure OpenAI (GPT-35-Turbo).
• Integrated a CV Skill Extraction Plugin that uses a RAG-powered Azure Function to parse user resumes and extract technical and domain-specific skills
• Developed a QueryRefinerPlugin using LLMs to clean, rephrase, and restructure user queries, improving downstream plugin accuracy and LLM understanding.
• Integrated external web search capabilities via the Tavily API to fetch real-time learning resources related to user skill gaps.
• Implemented Redis caching mechanism at the agent level to avoid repeated calls for the same questions, reducing token usage and API latency.
• Instrumented OpenTelemetry-based observability to trace token usage, model latency, and Azure OpenAI cost metrics across plugin calls and agent executions.
• Orchestrated plugin-based flow that chains refine → retrieve → search operations, returning structured responses like:
• Developed a FastAPI endpoint (/api/agtchat) to serve frontend interfaces like Streamlit, enabling real-time interaction with the assistant.

AI Engineer

Feb'22 - Mar2025

Elastacloud Limited, UK

• Intelligent Document Understanding & Conversational AI Platform
• Built an intelligent document pipeline using Azure Form Recognizer for OCR and schema extraction, integrated into a structured document store.
• Implemented RAG pipelines for context-aware answering, using LangChain and Semantic Kernel for multi-step reasoning and memory integration.
• Developed and deployed a FastAPI-based backend with Azure Functions for orchestration and summarization tasks
• DIntegrated LLM observability features (latency, response scoring, success/failure rates) using Azure Monitor and App Insights
• Deployed solution in a secure, scalable Azure environment with role-based access and integration into enterprise authentication
• Achieved 40%+ improvement in document retrieval accuracy by optimizing chunking, embedding models, and vector search configuration
• Optimised latency and throughput of AI Models for stable AI application
• Outcome: Enabled users to quickly understand and extract insights from complex business documents, significantly reducing manual review time and improving knowledge accessibility across the organization.

AI/Data Engineer

Jun'21 - Jan'22

BP(British Petroleum), UK

Cloud-Native Lakehouse & ML-Driven Analytics Platform

• Architected and built a modern Lakehouse solution on Azure using Databricks with Delta Lake for bronze/silver/gold data layering, supporting real-time and batch analytics pipelines.
• Automated ELT/ETL pipelines using Azure Data Factory, Azure Stream Analytics, and EventHub, enabling ingestion and processing of large-scale streaming security logs.
• Developed ML-enabled pipelines in Databricks notebooks for pattern detection, data profiling, and tag-matching—validated with 90%+ accuracy against benchmark datasets.
• Integrated Spark-based data quality and profiling workflows using PySpark, improving downstream ML readiness and reducing data noise.
• Collaborated with the AWS data team to support Data Mesh architecture using Lambda, S3, SQS, and DynamoDB for hybrid cloud processing and governance.
• Enabled downstream integration with Azure Sentinel for faster threat detection and alerting by transforming and standardizing high-volume log data.
• Outcome: Delivered BP’s first successful cloud-native data lake for SDS, reducing processing latency by 35% and enabling faster, more accurate cybersecurity response workflows across the organization.

Lead Consultant

Dec'18 - May'20

Genpact, India

Risk Management Platform Modernization

• Refactored and migrated monolithic risk management modules into modular Spring Boot microservices, reducing legacy dependencies and improving scalability
• Implemented Spring Batch workflows to manage large-scale data processing and schema migration, enabling seamless upgrades without data loss.
• Developed secure and efficient RESTful APIs to support inter-service communication and external system integration, resulting in a 30% performance improvement.
• Played a key role in code reviews, technical design discussions, and onboarding junior developers as part of the agile delivery team.
• Supported continuous deployment and testing through integration with CI/CD pipelines, ensuring consistent delivery in a regulated enterprise environment.

Senior Software Engineer

Dec'18 - May'20

RCS Tech, India

Enterprise Asset Management & Claims Processing Systems

• Designed and implemented core modules for an enterprise asset management system, deployed across GE and EDS, improving asset tracking and maintenance operations.
• Enhanced LDAP authentication and Single Sign-On (SSO) integration for enterprise applications, streamlining access management and user security.
• Developed and maintained Java-based web services to automate processes such as healthcare claims processing and employee tracking for Vidal Healthcare and HGS.
• Optimized backend performance by refactoring legacy code and improving database access layers using JDBC and Oracle stored procedures.
• Worked closely with client-side teams during on-site deployments and provided ongoing support to ensure smooth adoption and performance tuning.
• Outcome: Successfully improved platform responsiveness and maintainability while reducing technical debt. Delivered a resilient architecture that enabled
• Outcome: Delivered stable and maintainable software solutions that improved operational efficiency and security. Played a key role in long-term client engagements through strong technical delivery and consistent support.