Project 1 : ROI Agent

• Built an AI-powered ROI chatbot and decision engine using FastAPI and Mistral LLM, extracting financial entities from natural language inputs and reducing manual ROI analysis effort by 60%

• Designed a deterministic ROI computation engine, modeling logistics optimization, inventory savings, and cost avoidance, enabling faster and more accurate financial decision-making

• Implemented a RAG-based architecture (LangChain + vector database) to process domain-specific documents, improving response relevance and reducing query resolution time by 50% for complex financial queries

Client: Little River Transportation

• Built a full-stack, cloud-ready Transportation Management System (TMS) using React, FastAPI, PostgreSQL, and AI-assisted development tools (Gemini CLI, Google Antigravity), enabling real-time shipper–carrier matching based on lane capacity improving operational efficiency by 30%

• Designed and implemented data pipelines and analytics infrastructure, integrating PostgreSQL with Apache Superset to deliver interactive dashboards for lane analysis, capacity tracking, and operational insights, reducing manual reporting effort by 60%

• Engineered a scalable, containerized system architecture using Docker Compose, integrating FastAPI and PostgreSQL with Auth0-based authentication, enabling secure multi-user access and high-performance data processing

Client: Madex Associates

• Led end-to-end migration of on-premise logistics systems and data infrastructure to Azure, creating a centralized virtual environment for multi-user access and improving scalability

• Migrated Access-based data to Azure SQL, building ADF pipelines with self-hosted integration runtime to ingest 80+ tables (1M+ records) via ODBC into a scalable cloud architecture

• Engineered data ingestion and transformation pipelines, implementing a lift (raw) to web (normalized) schema and full-refresh strategy, improving data consistency and reducing data discrepancies by 50%

Project 1: The company relied on manual maintenance request creation, so we built an AI system to automatically generate requests from images and streamline the reporting process

• Developed a multi-user AI chatbot using GPT-4o to automate maintenance request generation from asset images, reducing manual reporting effort by 60%

• Built LLM-powered multimodal pipelines to process asset images and generate structured maintenance tasks, automating prioritization workflows and reducing processing time by 30%

• Implemented QR code-based asset tracking and role-based workflows, improving asset identification and enabling efficient multi-user collaboration across maintenance teams

Project 2: Users faced frequent heavy vehicle issues and struggled to find reliable solutions across scattered sources. Built a centralized AI system to deliver fast, accurate maintenance answers

• Scraped and processed 100K+ JSON threads and OEM maintenance documents, using Selenium, building a structured dataset of maintenance issues, solutions and preventive guidelines

• Designed data preprocessing pipelines (Pandas, NLTK) to clean and structure raw data into standardized JSON formats

• Built a RAG-based system using LangChain, embeddings, and vector databases (FAISS/Pinecone) to retrieve relevant context from large-scale datasets for accurate query resolution

• Developed a centralized AI platform enabling real-time, context-aware answers for heavy vehicle maintenance queries

Project 3: Built a predictive monitoring system to detect sensor failures early and enable proactive maintenance decisions

• Built a sensor failure analysis system by scraping and aggregating industrial sensor data (Siemens, Honeywell etc), creating structured datasets (10K+ records) for anomaly detection and predictive modeling

• Designed anomaly detection and pattern analysis workflows on real-time and historical data, improving early failure detection by 30% and reducing unexpected breakdowns

• Built and deployed fraud detection models using anomaly detection and classification techniques (Python, scikit-learn), reducing false positives by 35% and improving detection accuracy

• Automated end-to-end ML pipelines across 4,000+ Linux servers using Ansible, Salt, and Chef, reducing manual deployment effort by 60% and improving system reliability

• Implemented CI/CD pipelines for ML model lifecycle management using Docker and DevOps workflows, accelerating production rollout by 40%

• Migrated fraud analytics workloads from on-prem VMware to GCP, optimizing infrastructure utilization and reducing operational costs by 30%

• Developed real-time anomaly monitoring systems using log-based analysis and Python scripting, enabling automated model retraining triggers and faster fraud detection

• Deployed and scaled ML inference APIs on GCP, enabling continuous fraud scoring and reducing response time to fraud incidents by 40%

• Secured ML infrastructure using IAM, encryption policies, and compliance automation, ensuring data protection, governance, and regulatory adherence

• Delivered cloud migration and infrastructure solutions for enterprise customers, optimizing scalability, reliability, and operational efficiency across large-scale systems

• Supported enterprise infrastructure across 4,000+ Linux servers spanning 4 data centers, troubleshooting critical issues and maintaining high system availability (>99.9%)

• Partnered with customer-facing teams to gather requirements and deliver scalable cloud and infrastructure solutions, improving client satisfaction and system performance

• Led migration of on-premise VMware vSphere environments to GCP, improving scalability and reducing infrastructure costs by 30%

• Automated patching, provisioning, and configuration using Ansible, Salt, Chef, and Rundeck, reducing manual effort by 60% and improving deployment consistency

• Implemented infrastructure monitoring and security controls, reducing system incidents by 30% and improving overall reliability

• Communicated technical solutions and infrastructure changes to stakeholders, enabling better decision-making and simplifying complex systems for business teams