Senior Data Engineer & AI Solutions Architect with 5+ years' experience delivering end-to-end data platforms and intelligent applications across fintech, SaaS, and industrial analytics. Expert in designing scalable data pipelines, building AI-powered systems and Agentic AI with LLMs and machine learning models, and deploying robust cloud-native solutions on AWS, Azure, and GCP.
Experience
-
Senior Data Engineer & AI Solutions Architect @Helixiora
Netherlands - Remote
Summary:
Senior data and AI lead delivering customer-facing projects end to end: scalable Azure and AWS data platforms, agentic AI and RAG, LLM fine-tuning and serving, full-stack SaaS, and secure cloud operations with strong stakeholder partnership and mentorship.
Responsibilities:
Client Delivery & Solution Leadership: Led customer-facing project delivery by partnering with stakeholders to design and deploy innovative, end-to-end solutions across Azure and AWS, translating business problems into scalable data, AI, and platform architectures.
Data Engineering: Designed and built large-scale Azure Databricks pipelines using PySpark and SQL to ingest, clean, and transform multi-source datasets (Blob, CosmosDB); implemented real-time streaming pipelines with Kafka; and delivered curated Delta Lake layers aligned with Kimball-style dimensional modeling and data warehouse best practices to power analytics and ML workloads.
Full-Stack SaaS: Led the end-to-end development and deployment of multiple production-grade AI SaaS platforms, serving millions of users. Built full-stack solutions using FastAPI, Next.js, and React Native with Expo for Android and iOS, integrating OAuth 2.0 authentication, PostgreSQL, and Stripe to deliver secure, scalable, cloud-native products.
AI Agents: Architected and deployed multi-stage agentic RAG, Chatbot and automation systems using OpenAI Agent SDK, LangChain, LangGraph, CrewAI, Pinecone, and custom Python pipelines (Slack, Google Drive), enabling accurate context retrieval and intelligent workflow automation.
Machine Learning Engineering: Built feature-ready datasets in Delta Lake, supported downstream ML model training, and optimized vector search & reranking pipelines, accelerating model iteration cycles and improving retrieval accuracy across enterprise LLM solutions.
LLM Fine-Tuning: Fine-tuned domain-specific LLMs using Unsloth for specialized business use cases, optimized model performance with LoRA, and deployed the resulting fine-tuned models into production AI workflows, significantly improving inference accuracy.
LLM Inference & Deployment: Deployed vLLM and llama.cpp for high-performance LLM serving, optimized for low latency and high-concurrency workloads.
Infrastructure, DevOps & Security: Designed and automated CI/CD pipelines deploying containerized applications to AWS ECS using Docker, Terraform, and GitHub Actions; strengthened platform reliability with load balancing and scalable microservices; and implemented Zero-Trust controls, SSL/TLS hardening, firewall rules, and secure API gateways to ensure robust authentication, authorization, and data protection across SaaS and data platforms.
Monitoring & Observability: Set up centralized logging, metrics, and dashboards using Prometheus and Grafana, enabling real-time performance monitoring, alerting, and proactive issue resolution across production environments.
Team Leadership & Mentorship: Led and mentored junior and mid-level engineers through code reviews, architectural guidance, use of AI tools, and best-practice coaching, raising team capability and ensuring high quality engineering delivery.
Python
PySpark
SQL
Databricks
Delta Lake
Kafka
FastAPI
Next.js
React Native
Expo
LangChain
LangGraph
CrewAI
Pinecone
PostgreSQL
MySQL
Docker
AWS
Azure
Terraform
GitHub Actions
Git
Prometheus
Grafana
Nginx
-
Senior Data Engineer @Vaultoro
Manchester, United Kingdom - On-Site
Summary:
Architected Kimball-style warehousing and scalable financial ETL on GCP; co-led a FastAPI savings product and analytics supporting security and growth.
Responsibilities:
Architected a Kimball-style star-schema data warehouse using Elasticsearch and BigQuery, enabling real-time KPI dashboards in Kibana and supporting data-driven decision making.
Led the design and implementation of multiple scalable ETL pipelines with Python, GCP Dataflow, Scrapy, and managed workflow orchestration with Apache Airflow, processing over 10 million financial data rows daily for real-time analytics.
Co-led the development of a secure and scalable FastAPI-based Savings Platform, delivered in 3 months and now processing $2M+ in monthly customer deposits.
Managed and optimized databases including MongoDB, PostgreSQL, Elasticsearch, and BigQuery by tuning queries, indexes, and partitions, resulting in significant performance improvements.
Conducted deep data analysis with Pandas to detect anomalies and identify potential fraud patterns, enhancing platform security.
Automated data validation workflows using Python scripts, ensuring pipeline integrity and achieving 100% uptime for critical microservices.
Deployed containerized applications with Docker and Kubernetes, ensuring high availability, scalability, and streamlined CI/CD operations.
Developed and implemented a Random Forest classification model to identify high-value clients during signup, enabling personalized onboarding experiences.
Python
Elasticsearch
BigQuery
Kibana
GCP
FastAPI
MongoDB
PostgreSQL
Apache Airflow
Pandas
Docker
Kubernetes
-
Data Analyst @SoftCrop IT
Guwahati, India - On-site
Summary:
Delivered actionable insights via Tableau dashboards and automated data collection processes.
Automated competitor data scraping (Scrapy) and ETL into MySQL, reducing manual labour.
Analyzed sales and geographic data to guide strategic fibre network expansion.
Python
Tableau
MySQL
Scrapy
Pandas
Professional Projects
Projects delivered at Helixiora and Vaultoro, spanning data platforms, analytics, AI applications, and mobile products.
Helixiora
Solo developer · end-to-end delivery
AI Mental Health Assistant Mobile App and Admin Dashboard
Built an AI assistant for mental health support with an iOS and Android app and a Next.js super admin dashboard for the client.
Outcome: Delivered the complete mobile assistant and client admin dashboard independently in one month.
My contribution
Designed and implemented the Expo mobile app, FastAPI backend, PostgreSQL database, and AI agent using the OpenAI Agents SDK.
Built restricted administrative controls and integrated PostHog to track product usage and user engagement.
Optimized navigation, keyboard handling, and message rendering across iOS and Android; deployed the backend and database with Docker on a VPS.
Expo
React Native
Next.js
OpenAI Agents SDK
FastAPI
PostgreSQL
PostHog
Docker
Full-stack development
Beweginc Employee Wellbeing and Social Impact Platform
Developed an employee mobile app and company admin dashboard connecting wellbeing activities and challenges with charitable campaigns.
Outcome: Employees could track personal and team contributions while administrators managed campaigns, rewards, and participation.
My contribution
Built the Expo employee app and Next.js dashboard on a shared Convex backend with company membership and role checks.
Implemented activity credits, campaign goals, contribution tracking, and reward rules that preserved historical awards and prevented duplicate completions.
Shared domain logic and design tokens in a TypeScript monorepo, added English and Dutch support, and coordinated Vercel, Convex, and Expo EAS releases.
Expo
React Native
Next.js
Convex
TypeScript
Turborepo
Vercel
Expo EAS
Architecture and development leadership
On Premises RAG and AI Agent Platform
Designed and led development of an AI platform for chatting with Slack, Confluence, Google Drive, and uploaded documentation using locally hosted models.
Outcome: Delivered an extensible agent platform that kept content, indexes, and inference within the client's controlled environment.
My contribution
Designed the FastAPI and Next.js application, OAuth source connectors, and LangChain retrieval and agent workflows with source references and permission-aware retrieval.
Orchestrated incremental document ingestion, local embeddings, and vector indexing with Dagster, retaining ingested content and AI processing on-premises.
Led llama.cpp model hosting and concurrency tuning, implemented Stripe subscriptions, and enabled user-configurable MCP tool integrations.
FastAPI
Next.js
LangChain
Dagster
llama.cpp
OAuth
Stripe
MCP
Solution design and pipeline implementation
IoT Sensor Data Pipeline Optimization
Moved a growing IoT CSV workload from Azure Functions to Azure Databricks to support fruit and vegetable ripeness-detection ML training.
Outcome: Replaced the timeout-prone processing path with a scalable pipeline and traceable, validated datasets for ML training.
My contribution
Built incremental PySpark ingestion from Azure Blob Storage using Auto Loader and durable checkpoints.
Implemented Bronze, Silver, and Gold Delta Lake layers for raw records, schema validation, deduplication, and curated training datasets.
Published datasets to Cosmos DB with deterministic identifiers, idempotent writes, and controlled throughput; tuned Spark partitions and batch sizes.
Azure Databricks
PySpark
Auto Loader
Delta Lake
Azure Blob Storage
Cosmos DB
Data integration and BI delivery
Hospitality Sales and Workforce Dashboard
Built a consolidated view of restaurant sales, employee working hours, and performance across hotel and restaurant locations.
Outcome: Enabled the owner to compare locations, inspect employee hours, and make informed staffing and scheduling decisions.
My contribution
Integrated HR, workforce, and order APIs using scheduled Azure Functions and Cosmos DB, aligning location identifiers across sources.
Modeled sales and worked-hours facts with shared date and location dimensions in Power BI and built DAX measures for sales, hours, and sales per worked hour.
Added incremental upserts, source reconciliation, Application Insights monitoring, and refresh coordination to keep reporting accurate and current.
Azure Functions
Cosmos DB
REST APIs
Power BI
Power Query
DAX
Application Insights
Model training, evaluation, and implementation
Inappropriate Language Filter with LLM Fine-Tuning
Fine-tuned a small open-source LLM on client data to classify inappropriate language in context using Unsloth and LoRA adapters.
Outcome: Reached around 98% overall accuracy with inference below one second and lower inference costs.
My contribution
Prepared labeled examples and separate training, validation, and held-out test sets, grouping related examples to reduce data leakage.
Tuned LoRA training settings and evaluated precision, recall, F1, false positives, and false negatives.
Integrated validated moderation outputs and versioned datasets, model artifacts, and evaluation results for repeatable updates and rollback.
Python
Unsloth
LoRA
Open-source LLM
Supervised fine-tuning
Vaultoro
End-to-end project ownership
BI and Analytics Platform
Created a centralized warehouse and BI platform so owners and board members could review business performance without requesting database reports from developers.
Outcome: Gave business stakeholders direct access to consolidated financial and customer reporting, reducing dependence on developer-prepared reports.
My contribution
Designed a BigQuery warehouse using Kimball star schemas and ingested application and financial data from MongoDB and PostgreSQL with Python.
Orchestrated pipelines with Apache Airflow on GCP and built Google Data Studio dashboards for profit margins, customers, and business performance.
Implemented incremental, idempotent loading with staging, deduplication, financial reconciliation, checkpoint transactions, and recoverable batches.
BigQuery
Python
Apache Airflow
GCP
MongoDB
PostgreSQL
Google Data Studio
Technical design and implementation
Fraud and Money Laundering Detection
Built daily monitoring of suspicious financial activity with explainable detection rules and a dashboard for investigation.
Outcome: Made suspicious patterns and supporting evidence visible in one dashboard so reviewers could prioritize and investigate potential financial crime.
My contribution
Analyzed historical transactions and worked with a financial expert to define rules, thresholds, and review priorities.
Implemented a configurable Python rule engine and an investigation dashboard showing supporting transactions, triggered rules, and review statuses.
Deployed with Docker on GCP and added deterministic alert keys, incremental checks, validation, monitoring, and replayable evaluation windows.
Python
Docker
GCP
Rule-based detection
Transaction analysis
Monitoring dashboards
Backend architecture, database design, and BI ownership
Investment App Backend and BI
Owned the shared backend and MongoDB data model for a mobile and web investment app, with a reporting system ready before launch.
Outcome: The team delivered the app in three months, and customers invested over $2 million through the platform.
My contribution
Built FastAPI endpoints for investments, transaction history, and status tracking with Pydantic validation and customer-scoped access.
Designed financial records with decimal precision, guarded status transitions, audit history, and idempotent investment requests.
Built incremental MongoDB-to-BigQuery pipelines and Google Data Studio dashboards, validating investment totals and transaction outcomes before launch.
A comprehensive collection of production-ready Docker Compose stacks for self-hosted services, including monitoring, databases, home automation, and infrastructure management tools for personal homelab environment.
🐳 Designed and deployed 15+ production-ready Docker Compose stacks for self-hosted services
📊 Implemented comprehensive monitoring stack with Prometheus, Grafana, cAdvisor, and Node Exporter for infrastructure visibility
🔍 Built Elasticsearch stack for log aggregation and search capabilities with Kibana visualization
🏠 Integrated Home Assistant for smart home automation and IoT device management
💾 Configured MySQL and database services with optimized Docker configurations for data persistence
🚀 Deployed Portainer for container management and orchestration with streamlined deployment workflows
⚡ Implemented performance testing tools (LibreSpeed, OpenSpeedTest) for network and system benchmarking
A Python automation script that automatically tags anime episodes as filler or canon by scraping animefillerlist.com and intelligently renaming files with metadata for better organization and viewing experience.
🔍 Built automated web scraper to extract episode metadata from animefillerlist.com for accurate filler/canon classification
📁 Implemented intelligent file renaming system that preserves quality tags while adding filler/canon metadata
⚙️ Developed configurable system supporting multiple anime series with customizable quality tags and file paths
🎯 Created smart episode detection algorithm to handle various filename formats and episode numbering schemes
📊 Automated metadata integration that enhances media library organization and viewing experience
🚀 Streamlined workflow for anime enthusiasts to efficiently manage large episode collections
A computer vision-based hand gesture recognition system using TensorFlow and OpenCV for real-time gaming control, enabling hands-free game interaction through custom-trained machine learning models.
🤖 Built custom hand gesture detection model using TensorFlow transfer learning with SSD MobileNet V2 FPNLite architecture
📸 Developed automated image collection pipeline for custom hand gesture dataset creation and labeling
🎯 Implemented real-time hand gesture recognition from video feed for gaming applications
🎮 Created practical gaming integration demonstrated with Chrome Dino game control
🔬 Applied transfer learning techniques to optimize model performance for specific gesture recognition tasks
📊 Structured project workflow with Jupyter notebooks for data collection, training, and detection phases
A Python-based automation tool that downloads and applies the latest Nvidia DLSS and DLSS Frame Generation DLLs to local games, ensuring optimal gaming performance and visual quality.
🔧 Built automated DLL management system for Nvidia DLSS and Frame Generation updates
🌐 Integrated with TechPowerUp.com API for real-time DLL version monitoring and downloads
📁 Implemented intelligent file discovery to locate DLSS DLLs across multiple game directories
💾 Created automated backup system with timestamped naming for safe DLL rollback
⚙️ Developed configurable system supporting multiple game library paths and server locations
🚀 Packaged as executable with PyInstaller for easy distribution and deployment
⭐ Achieved 10+ stars on GitHub demonstrating community recognition and utility
A Python-based web scraping application that monitors Gameloot.in for PC component stock changes and sends real-time Telegram notifications when products become available or go out of stock.
🕷️ Built automated web scraper using BeautifulSoup and Python for real-time PC component monitoring
📊 Implemented MongoDB integration for persistent storage and smart deduplication of products
🤖 Developed Telegram bot integration for instant stock change notifications
🔍 Implemented change detection algorithms to identify new products, restocked items, and sold-out products
📝 Comprehensive logging system with configurable levels for monitoring and debugging
🚀 Automated deployment with error handling and retry mechanisms for robust operation
ETA Prediction for Container Ports using Machine Learning
MSc project at University of Liverpool predicting ship delays with 80% accuracy using machine learning techniques.
🎯 Predicted ship delays 10+ days in advance with 80% accuracy
🔧 Cleaned and engineered features using Pandas, NumPy, and Pearson correlation
🤖 Evaluated multiple ML models (SVM, DT, RF, NN) and selected Random Forest
🚀 Deployed the model with FastAPI for real-time predictions
Key Modules: Data Mining & Machine Learning, Big Data Analytics, Enterprise Systems with SAP, Digital Business Technology and Management, Digital Strategy
