Building Machine Learning Models & Practical AI Systems
I build and experiment with machine learning, generative AI, data analytics, and data engineering solutions. My projects focus on turning real-world problems into practical applications while continuously learning, experimenting, and improving my technical skills.
“I build and study practical AI/ML systems, data products, and intelligent applications.”
I'm a Data Science and AI/ML practitioner focused on building practical solutions with data, machine learning, and modern AI technologies. I've worked on projects spanning predictive modeling, data analytics, generative AI, RAG applications, and data engineering, using these projects to turn what I learn into working applications.
I enjoy understanding how systems work end-to-end—from exploring and preparing data to developing models, building APIs and applications, and experimenting with deployment and MLOps workflows. I'm continuously expanding my skills through hands-on projects, experimentation, and learning new technologies.
Building and evaluating machine learning models for classification, regression, prediction, and real-world data problems, with a focus on experimentation, evaluation, and improvement.
Exploring and building LLM-powered applications, retrieval-augmented generation systems, and AI workflows using modern language models, embeddings, and vector databases.
Analyzing real-world datasets, uncovering patterns, and building predictive models and interactive dashboards that turn data into useful insights.
Working with data pipelines, SQL, APIs, Docker, cloud platforms, and deployment workflows to understand how data and ML applications move from development toward usable systems.
Tools and technologies I work with every day
Areas I'm actively studying and building expertise in right now.
Building autonomous AI agents with tool use, planning, and memory using LangGraph and CrewAI frameworks.
Deploying and scaling ML workloads on Kubernetes clusters using Kubeflow and custom operators.
Fine-tuning large language models efficiently using Parameter-Efficient Fine-Tuning techniques on domain-specific datasets.
Learning Rust to build high-performance data processing tools and extend Python ML pipelines with native extensions.
Studying GNNs for fraud detection, recommendation systems, and knowledge graph construction using PyG.
Implementing data contracts, schema validation, and automated quality checks in production data pipelines.
Hands-on projects that demonstrate my skills in machine learning, generative AI, data analytics, and real-world problem solving.
Extracts, parses, and analyzes financial documents (PDFs, reports, statements) using NLP and LLMs with a RAG-based QA system and structured citation retrieval.
High-throughput microservice pipeline leveraging fine-tuned Transformer models for multi-task text classification, entity extraction, and multi-modal image synthesis.
Forecasts demand, optimizes inventory, and recommends purchase orders using machine learning and time series analysis to reduce kitchen waste and improve replenishment.
Explore real-time repositories and open-source applications in Machine Learning, AI, Data Engineering, and Web Development. Click any card to explore details.
End-to-end application leveraging Hugging Face pretrained models for multi-task NLP and image generation (Streamlit/Gradio front-end, model evaluation).
Comparative benchmark of CNN architectures (LeNet, AlexNet, GoogLeNet, ResNet, Xception) on MNIST / Fashion-MNIST / CIFAR-10 with full evaluation metrics.
PyTorch sequence-to-sequence demo exploring attention mechanisms (synthetic reverse-sequence task) with training scripts and visualizations.
Classification pipeline for cybersecurity incidents (uses GUIDE dataset) to assist SOC triage — includes modeling, evaluation and analysis.
Sales forecasting project for Domino's ingredient ordering — historical-sales forecasting to optimize orders and reduce waste.
Streamlit dashboard + Selenium scraper that collects RedBus travel data and visualizes trends for operational analysis.
An AI-powered Streamlit dashboard that extracts, parses, and analyzes financial documents (PDFs, reports, statements) to provide structured insights and visualizations for decision-making.
A data storytelling dashboard that analyzes global electronics market trends, supply chain metrics, and operational KPIs, offering interactive insights for strategic planning.
Professional certifications and credentials that validate my expertise.
DeepLearning.AI
Stanford University / Coursera
Amazon Web Services
IBM / Coursera
Google / Coursera
Activeloop
Real-time AI research from the world's leading labs
Sep 25, 2026 • openai.com
With Codex, GPT-Live-1, and GPT-6 Astra, Proaction builds, operates, and sells modern fleet management faster.
Sep 23, 2026 • openai.com
Marking two years of OpenAI Academy and bringing AI skills to even more communities.
Sep 23, 2026 • openai.com
OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure.
I'm available for collaboration, consulting, or tackling data science challenges. Whether it's a project, an idea, or a new opportunity — I'd love to hear from you.
Insights into engineering methodology, model optimization techniques, and business impact strategy.
Specializing in end-to-end machine learning engineering, MLOps implementation, and custom Generative AI systems. This spans data architecture, model development, and deploying scalable inference pipelines on AWS or GCP.
Following a structured MLOps workflow: aligning business objectives, performing rigorous exploratory data analysis (EDA), training baseline models, and deploying containerized solutions equipped with automated data drift monitoring.
Building scalable Retrieval-Augmented Generation systems using vector stores like Pinecone or Qdrant, implementing hybrid search (dense + sparse) and re-ranking algorithms to eliminate hallucinations and secure query responses.
Selected based on the use case: RAG is optimal for dynamic, external knowledge retrieval with strict data freshness, whereas Fine-Tuning is chosen when modifying a model's specialized style, tone, or domain-specific grammar.
Utilizing model quantization (INT8/FP16), structured pruning, and hardware acceleration frameworks like ONNX Runtime or TensorRT to reduce memory footprint and latency while retaining predictive precision.
Defining baseline metrics before development, optimizing cost-per-inference ratios, and executing A/B testing frameworks to measure direct business impacts on conversion rates, automation throughput, and operational costs.
