Our Portfolio
Production-grade AI systems, Generative AI applications, distributed platforms, and enterprise integrations — shipped and running at scale.
AI Voice Bot Platform
We are actively building a production-grade AI Voice Bot platform that goes far beyond basic IVR. The system connects real-time Speech-to-Text, a frontier LLM reasoning layer, and low-latency Text-to-Speech in a streaming pipeline over WebRTC — enabling natural, context-aware conversations at scale. The platform will support inbound and outbound call automation, live agent assist, CRM integration, and multi-language conversations, with a dashboard for analytics and call management.
Targeting a 10× reduction in call centre costs through human-like AI voice agents handling thousands of concurrent calls.
Enterprise IT Service Desk AI — MCP Agent & Teams Bot
Designed and deployed two tightly integrated components for a confidential enterprise client. The first is a production MCP server (FastMCP 2.0) connecting Claude to ServiceNow, Jira, and Confluence through 24 purpose-built tools — covering parallel ticket research, CMDB dependency mapping, change-to-incident correlation, on-call routing, and a self-improving quality tagging layer that surfaces proven resolutions over time. The second is a Microsoft Teams bot (FastAPI + GPT-4o) handling ~20 ITSM intents in real time: ticket creation, AI auto-triage, status checks, software access requests, and RDP session termination. A nightly background service syncs and vectorizes Confluence KB articles into FAISS and pgvector for semantic knowledge retrieval. Both components share Okta OAuth 2.0 auth, Active Directory group gating, CyberArk Conjur secrets, and run on AWS EKS via Helm. CI/CD via CircleCI → Octopus Deploy across five environments.
Replaced hours of manual cross-system research per shift for Service Desk teams and automated Level-1 triage and knowledge retrieval for all enterprise users directly inside Teams.
AI-Powered Candidate Screening & Evaluation Platform
Designed and deployed a production-grade AI recruitment platform to automate high-volume candidate screening for enterprise hiring teams. The system uses transformer-based NLP pipelines for deep resume understanding, LLM-based contextual scoring and ranking, and a microservices backend built on FastAPI with horizontally scalable APIs. The platform integrates with existing ATS systems and surfaces structured insights to hiring managers, dramatically cutting time-to-shortlist.
Reduced manual screening effort by over 70% and significantly improved hiring accuracy at scale.
Enterprise Knowledge Copilot (RAG-based AI System)
Built a context-aware AI copilot that enables employees to query and interact with vast enterprise knowledge bases in natural language. The system uses a RAG architecture with vector embeddings and semantic search for precise retrieval, integrates natively with Confluence, ServiceNow, and internal documentation systems, and orchestrates responses via LangChain with FAISS and Pinecone for vector storage. Deployed on Kubernetes on AWS with robust access controls and audit logging.
Enabled real-time enterprise knowledge access for 1,000+ users, reducing support ticket volume and onboarding time.
Large-Scale Image Comparison & Computer Vision Pipeline
Architected a distributed computer vision pipeline capable of processing millions of images for similarity comparison, deduplication, and visual search. The system employs CNN-based deep feature extraction and embedding generation, with GPU-optimised inference pipelines built on TensorFlow and PyTorch. Asynchronous job queues and auto-scaling workers on AWS handle burst workloads, while OpenCV powers pre-processing and post-processing stages. Results are indexed for fast approximate nearest-neighbour search.
Reduced image analysis time from hours to minutes at production volume, enabling real-time visual search capabilities.
Event-Driven Distributed Platform for Real-Time Systems
Designed and built a highly resilient, event-driven distributed platform to power real-time data processing for high-throughput business workloads. The platform uses a microservices architecture with Spring Boot and Node.js services communicating over Kafka topics and AWS SQS/SNS. Complex multi-step business workflows are orchestrated reliably using Temporal, with automatic retries, timeouts, and visibility into workflow state. The entire stack runs on Kubernetes with horizontal auto-scaling and blue-green deployment support.
Achieved reliable, low-latency processing of millions of real-time events per day with 99.9% uptime.
Intelligent Business Process Automation Platform
Designed and deployed a business process automation platform that replaced hundreds of hours of manual cross-department work each month. Using n8n as the core orchestration engine, the platform connects CRM, HRMS, finance, and communication tools through intelligent workflows triggered by business events. AI classification and extraction layers built on LLMs pre-process incoming data — emails, forms, documents — before routing them through the correct automation path. Custom webhook integrations, retry logic, and audit trails ensure reliability and compliance. The platform was deployed on a self-hosted Kubernetes cluster for full data sovereignty.
Eliminated over 300 hours of manual work per month across 4 departments, with zero data loss and full audit compliance.
Enterprise AI Integration & Digital Transformation Layer
Designed and delivered a transformation layer that brings modern AI capabilities into complex, legacy enterprise environments without disrupting existing systems. The solution integrates AI-powered automation into SAP S/4HANA and NetWeaver landscapes via secure OData and REST APIs, using SAP UI5/Fiori for the frontend experience. Deployed on OpenShift with AWS and Azure support, the platform enforces enterprise-grade security, SSO, and audit compliance throughout.
Accelerated AI adoption across enterprise teams without disrupting existing SAP workflows or compliance posture.