Skills
Machine Learning Algorithms: Supervised & Unsupervised Learning, Model Stacking, Hyperparameter Tuning.
Deep Learning: Deep Learning Architectures, Model Optimization, Model Evaluation.
Statistical Foundation: Calculus, Probability, and Mathematical Modeling for ML Algorithms.
Experimentation: Experiment-driven model development and performance benchmarking.
LLM Fine-Tuning: QLoRA, SFT (Supervised Fine-Tuning), Unsloth Optimization.
RAG Architecture: End-to-end Retrieval-Augmented Generation, Multi-stage RAG Pipelines.
Advanced Prompt Engineering: Multi-step Reasoning, Chain-of-Thought, Structured Output Generation.
Agentic Workflows: Autonomous LLM Agents, Multi-agent Task Orchestration.
Text Analytics: Sentiment Analysis, Comment Classification, Text Cleaning Pipelines.
Domain Adaptation: Fine-tuning models for specific domains (e.g., Legal, Business).
Object Detection: YOLO (You Only Look Once) framework for real-time detection.
Document Intelligence: OCR-based processing, Structural & Semantic Document Analysis.
Image Processing: Feature extraction and automated label detection.
Vector Databases: Qdrant, Semantic Search, Similarity Search.
Advanced Retrieval: BGE-M3 Embedding, Multi-vector Retrieval, Cross-encoders, Reranking Strategies.
Data Processing: Advanced Chunking (Structural, Semantic, Sliding Window), Dataset Automation.
Dataset Engineering: Instruction-formatted dataset construction, Dataset Cleaning & Augmentation.
Programming Languages: Python, Go, SQL, JavaScript, C, C++.
Frameworks & Libraries: PyTorch, TensorFlow, LangChain, LlamaIndex, Scikit-Learn.
Specialized AI Tools: Ollama, Unsloth, PyMuPDF, PyMuPDF4LLM.
System & Deployment: API Development, Model Serving, Docker, Cloud Training (GCP).
Mobile Development: Flutter (Cross-platform integration for AI models).
Model Validation: Citation validation mechanisms, traceable output verification.
System Architecture: Designing scalable AI system architectures and end-to-end data pipelines.
About
Results-driven AI Engineer and Founder of STRATA AI with a proven track record in architecting end-to-end AI pipelines and scalable document intelligence systems. Expert in Natural Language Processing (NLP) and Retrieval-Augmented Generation (RAG), evidenced by the development of a multi-stage retrieval framework using Qdrant and BGE-M3 embeddings, featuring advanced structural chunking and citation validation mechanisms for high-stakes legal documentation. Highly proficient in Deep Learning and Model Optimization, successfully fine-tuning domain-specific LLMs like Qwen3 3B using QLoRA and Unsloth to reduce training resource requirements while maintaining superior reasoning capabilities.In the field of Computer Vision, served as Technical Lead for the Sadar-Gizi application, integrating YOLO-based object detection models for automated nutritional label extraction and real-time inference. Demonstrated expertise in Machine Learning and text analytics as a Top 20 Finalist in the IBM x Hacktiv8 competition, building an automated sentiment analysis system using IBM Granite LLM to generate insights from high-volume social media data. Technical stack includes Python, Go, PyTorch, TensorFlow, and LangChain, with a focus on building robust, experiment-driven AI solutions that bridge the gap between research and practical implementation.