MD ALI ARMAGHAN

AI-ML Engineer
New Delhi, IN.

About

AI Engineer with over 2 years of experience in developing, fine-tuning, and deploying machine learning models, with a strong focus on Generative AI, LLMs, and deep learning. Possessing expertise across the end-to-end AI model lifecycle, including data preprocessing, training, hyperparameter tuning, fine-tuning (LoRA, QLoRA), and production deployment. Proficient in deep learning frameworks (PyTorch, TensorFlow), LLM optimization (RLHF, DPO), vector databases (FAISS, ChromaDB), and cloud-based AI deployment. Driven by a passion for improving model efficiency, reducing hallucinations, and advancing AI applications through cutting-edge research.

Work

Maya AI
|

Prompt Engineer

Summary

Led the development and deployment of advanced LLM solutions for banking automation, focusing on improving accuracy and efficiency.

Highlights

Developed, fine-tuned, and deployed LLMs for banking automation, improving AI response accuracy by 30%.

Built and optimized custom RAG pipelines using FAISS and ChromaDB, significantly enhancing retrieval efficiency.

Trained and fine-tuned transformer models (LLaMA, GPT) using LoRA and QLoRA, resulting in reduced model memory usage.

Integrated AI models with FastAPI and deployed on AWS SageMaker for seamless real-time interactions.

Conducted comprehensive LLM evaluations using perplexity scores, BLEU, ROUGE, and human feedback loops to drive continuous improvement.

Remotasks
|

AI Engineer | LLM Optimization Specialist

Summary

Focused on engineering and optimizing AI-powered code generation models, ensuring high accuracy and reliability.

Highlights

Engineered and optimized AI-powered code generation models, increasing accuracy by 50%.

Fine-tuned models for specialized tasks including code completion, bug detection, and code refactoring.

Implemented Reinforcement Learning from Human Feedback (RLHF) for model evaluation, ensuring correctness and reducing bias in AI-generated code.

Automated model training and deployment workflows utilizing Hugging Face Spaces and AWS Lambda.

Pixel Flame Software pvt. ltd.
|

Junior Data Scientist

Summary

Contributed to data-driven solutions, including developing predictive models and enhancing recruitment processes through AI.

Highlights

Developed and deployed a salary prediction model using Scikit-learn, improving accuracy by 15%.

Processed and analyzed over 100,000 employee records using Pandas and SQL, applying robust feature engineering techniques.

Built a resume screening system utilizing NLP techniques (TF-IDF, Word2Vec), significantly improving recruitment efficiency.

Collaborated with engineering teams to deploy AI models and meticulously documented findings for continuous improvement.

Education

Aligarh Muslim University

Bachelor of Technology

Computer Science

Grade: 8.028/10

Languages

English

Fluent

Certificates

Generative AI with LangChain and HuggingFace

Issued By

Udemy

LLMOps

Issued By

DeepLearning.AI

CS-50's Introduction to Artificial Intelligence with Python

Issued By

Edx-Harvard Online

Skills

Programming Languages

Python.

Machine Learning Frameworks

PyTorch, TensorFlow, Scikit-learn, HuggingFace Transformers.

Machine Learning Concepts

Supervised Learning, Unsupervised Learning, Deep Learning, Transfer Learning, Reinforcement Learning.

LLM Optimization

Fine-tuning (LoRA, QLoRA, PEFT), Model Distillation, Quantization (GPTQ, AWQ), Prompt Engineering, RLHF, DPO.

Generative AI Models

OpenAI, Meta AI (LLaMA), Groq, Mistral, Stable Diffusion, SpeechT5, Nvidia NIM, Crew AI.

Databases & Deployment

FAISS, ChromaDB, AstraDB, Weaviate, SQL Server, MySQL, Git, HuggingFace Spaces, AWS SageMaker, AWS Lambda, FastAPI, Streamlit, Triton Inference Server.

NLP & Data Processing

LangChain, Pandas, SQL, TF-IDF, Word2Vec, Hugging Face Datasets, Data Augmentation, Natural Language Processing, Sentiment Analysis.

Projects

Intelligent Customer Support Agent

Summary

Developed an AI-driven customer support agent designed to automate query categorization, sentiment analysis, and response generation, leveraging advanced NLP and generative AI capabilities.

Fine-Tuning DeepSeek-R1 for a Medical Chatbot

Summary

Fine-tuned the DeepSeek-R1 model for medical Q&A tasks, focusing on optimizing model efficiency and enhancing response accuracy for healthcare applications.