Hello World, I'm Sadjad, Data Scientist | Machine Learning/MLOps Engineer | Computational Neuroscientist (Ph.D.)! 👋
- Data Scientist and MLOps Engineer with 8+ years in machine learning, data analysis, and brain simulations.
- Proficient in Python, TensorFlow, PyTorch, Keras, and computer vision (CNNs, OpenCV).
- Developed AI/ML solutions, including a real-time blood cell classification system.
- Skilled in Docker, Kubernetes, AWS, and Azure with CI/CD automation (Git, GitHub Actions, Azure DevOps).
- Expertise in SQL, NoSQL (MongoDB), and data visualization (Matplotlib, Seaborn, Plotly).
- Experience with cloud monitoring using Prometheus and Grafana.
- Published research applying machine learning to fMRI data and Bayesian modeling.
- Developed APIs and deployment pipelines with Flask, FastAPI, Docker, and Terraform.
- Passionate about using AI/ML to solve complex problems in healthcare and finance.
- Open to collaborations on projects related to data science, MLOps, AI/ML pipelines.
Machine Learning Engineer Diploma |
Data Science Diploma |
Azure Fundamentals |
| Projects | Techniques | Data Types | Poster |
|---|---|---|---|
|
Self-Improving LLM Knowledge Base Production-grade RAG system with hybrid retrieval (FAISS + BM25), persistent self-improving memory that learns from user interactions, and a feedback loop that automatically generates and integrates new knowledge summaries back into the knowledge base. |
Retrieval-Augmented Generation (RAG), Hybrid Retrieval (FAISS Dense Vectors + BM25 Sparse), Reciprocal Rank Fusion (RRF), Semantic Chunking, Grounded LLM Reasoning, Self-Improving Memory Loop, Persistent Knowledge Store, Importance Scoring, Query Deduplication, Semantic Similarity Search, Anti-Hallucination Prompting, LLM-as-Judge Evaluation, Experiment Tracking (MLflow), Fine-grained Architecture Design, Context Engineering | Markdown Documents, Text Queries, User Questions, Natural Language Input, Configuration Files (YAML), API Responses, Chat Messages, Structured Metadata |  |
|
Predictive Maintenance MLOps End-to-end predictive maintenance system using machine learning and MLOps principles. Includes failure prediction, remaining useful life estimation, anomaly detection, and interactive Power BI & Tableau dashboards for industrial decision-making. |
Failure Classification, Remaining Useful Life (RUL) Regression, Anomaly Detection, Time-Series Modeling, Ensemble Methods, XGBoost, LightGBM, LSTM, Isolation Forest, Autoencoder, Feature Engineering, SHAP Explainability, Cost-Sensitive Learning, What-If Simulation, MLOps Orchestration | IoT Sensor Data (CSV/Parquet), Real-Time Sensor Readings (Temperature, Vibration, Pressure, Rotation Speed), Historical Sensor Telemetry, Maintenance Logs, REST API Payloads (JSON) |  |
|
RecSys ML Platform Production-grade real-time recommendation system with microservices, Kafka/Spark pipelines, MLflow lifecycle management, online learning, A/B testing, and Prometheus/Grafana monitoring, deployed via Docker and Kubernetes. |
Microservices Architecture, Collaborative Filtering (ALS), Online Learning (Incremental Embeddings), A/B Testing (Deterministic Hashing), Streaming Data Processing (Kafka + Spark), Feature Engineering, Model Serving, FastAPI, MLflow Registry, Prometheus/Grafana Monitoring, Docker/Kubernetes Orchestration, Redis Caching, Delta Lake, Data Drift Detection (Evidently) | Clickstream Events (clicks, views, ratings), User Interaction Payloads, JSON API Requests, Kafka Event Streams |  |
|
Enterprise RAG Assistant GDPR-compliant, agentic RAG system with LangGraph, hybrid LLMs (Ollama + GPT), and enterprise document retrieval. |
LLM (Ollama + GPT-4o-mini), LangChain, LangGraph, Agentic Workflow, RAG, Hybrid Search (Vector + BM25), Prompt Engineering, GDPR Compliance, PII Anonymization, Multilingual NLP, FastAPI, Weaviate Vector DB, Document Processing, Token Counting, Language Detection | Text, PDF, Markdown, CSV/Structured Documents, JSON, API Data, Embeddings, User Queries, Cost Metrics, API Responses, Chat Messages, Document Metadata |  |
|
Open Autonomous Research Lab (OARL) A multi-agent AI platform for autonomous data analysis, ML experimentation, research discovery, and report generation. |
Multi-Agent System, Agentic Workflow, Orchestration Agents, Ensemble Learning (XGBoost, LightGBM), Statistical Analysis (SciPy, Statsmodels), Vector Database (ChromaDB), RAG (Retrieval-Augmented Generation), Experiment Tracking (MLflow), MCP Servers, Prompt Engineering, FastAPI REST Architecture, Streamlit UI | CSV, Pandas DataFrames, NumPy Arrays, JSON, PyArrow Tables, SQL Databases, Text Reports, Experiment Metadata, Vector Embeddings, Excel Files, Tabular Data, Time Series Data |  |
|
AegisClaims AI AegisClaims AI is a production-grade, multi-tenant B2B SaaS platform that provides AI-powered, autonomous insurance claims triage and decisioning for motor and property insurance. |
LLM (AWS Bedrock), Multi-Agent AI System, RAG (Retrieval-Augmented Generation via OpenSearch), ML-based Fraud Detection (AWS SageMaker), Clean Architecture, Agentic Orchestration, NLP, OCR, RBAC, Multi-Tenancy, Audit Trails, Prompt Engineering | Claims Data, Insurance Policy Documents, Unstructured Documents (OCR), Claim Metadata, Fraud Detection Signals, User Profile Data, Tenant Configuration Data, Audit Logs, Prompt Templates, Model Evaluation Datasets |  |
|
RL Environment Framework Production-grade deterministic RL environment framework for training and evaluating LLM-style agents without external APIs. |
Reinforcement Learning (RL), PPO (Proximal Policy Optimization), Generalized Advantage Estimation (GAE), Deterministic Environment Simulation, Out-Of-Distribution (OOD) Evaluation, Distribution Shift Analysis, Deterministic Judging, Policy Gradient Methods, Stability Instrumentation, YAML Configuration, PyTorch, Docker, Pytest, GitHub Actions, CI/CD Automation | Synthetic Datasets, Feature Vectors (32-dimensional), Classification Labels (5 classes), Train/Validation Splits, Policy Checkpoints, Metrics Logs (JSON), YAML Configurations, Covariance Matrices, Class Priors, Reward Signals, Gradient Data, Entropy Values, KL Divergence Metrics, Accuracy Scores, Generalization Gap Data |  |
|
FinSage: AI-Powered Financial Advisor Autonomous multi-agent system for financial and investment advising, helping users make data-driven decisions across various asset classes including stocks, ETFs, bonds, and cryptocurrencies. |
LLM (Gemma3 via Ollama), LangChain, LangGraph, Agentic Workflow, FinBERT, RAG, Prompt Engineering, MCP (Modular Contextual Pipeline), Gradio UI | Text, CSV, JSON, API Data, News Articles, User Profile Data, Scheduling Settings |  |
|
Personal Knowledge Management Agent Context-aware Q&A over personal notes (Markdown, Notion exports, Obsidian vaults), with daily summarization and learning digest via autonomous agents. |
LLM (Gemma3 via Ollama), LlamaIndex, Qdrant, RAG, Agentic Scheduling, Prompt Engineering, LangChain | Text, Markdown, Notion/Obsidian Exports |  |
|
PDF Q&A RAG DeepSeek-R1 PDF Q&A Assistant. |
LLM (DeepSeek-R1), HuggingFace Embeddings, Vector Search, Sentence Splitting, Prompt Engineering, Streaming Responses | Text, PDFs |  |
|
Text-Summarizer Pegasus Text Summarization. |
Transformers (Pegasus), Fine-Tuning, Data Augmentation, Tokenization, Beam Search, Length Penalty, CI/CD (GitHub Actions), FastAPI, Docker, AWS | Text (Dialogues, Summaries) |  |
|
Malaria Cell Classifier Deep Convolutional Neural Networks and Machine Learning Models for Anomaly Detection in Microscopic Malaria cells. |
Deep CNN, Data Augmentation, Feature Engineering, Image Processing, Optimization | Image |  |
|
Automated Nucleus Detector A Semantic Segmentation Solution for Automating Nucleus Detection of Microscopic Biomedical Images. |
U-Net, Keras-tuner, Semantic Segmentation | Image |  |
|
Movie Recommendation Systems A comprehensive collection of movie recommendation systems, implementing collaborative filtering, content-based filtering, and Bayesian average techniques. |
Collaborative Filtering, Content-Based Filtering, Bayesian Average | Metadata, User Ratings, CSV |  |