Hi, I'm Aquib A. Khan š
Data Scientist & Machine Learning Engineer with experience in building scalable AI systems, and deploying ML in production. Currently pursuing M.S. in Data Science at Pace University, New York.
Ask the chatbot anything about me
Click the little icon in the top bar to activate the AI chat. You can ask the chatbot any question about me, my work experience, projects, and skills.
About Me
Who am I?
My name is Aquib Ali Khan. I am a Data Scientist & ML Engineer with 3+ years of experience, specializing in shipping scalable AI solutions. My expertise spans classical ML, NLP, computer vision, and end-to-end production deployment, from parameter tunning to building robust retrieval pipelines for industry.
I'm passionate about deploying scalable NLP, CV, and classical ML models to production, ensuring robust and efficient AI solutions.
Professional Experience
AI Engineer Intern ā National Institute of Technology (Remote)
Sep 2023 ā Sep 2024
- Led development of scalable RAG systems using LangChain, VectorDB, and advanced prompting, enabling production-grade LLM outputs with higher factual accuracy and contextual relevance.
- Scraped and preprocessed large-scale domain-specific data using automated pipelines, enabling high-quality embeddings and up-to-date knowledge integration for LLMs.
- Architected hybrid RAG pipelines, quantized 3M+ vectors, and integrated PostgreSQL for preprocessing, reducing embedding generation time.
- Applied multi-query, decomposition, Step-Back, and HyDE prompting techniques, plus logical and semantic routing with Pydantic schemas, boosting document recall and routing precision.
- Fine-tuned LLM APIs (o4-mini, Claude 3 Haiku) and open-source models (LLaMA, Phi, Qwen) using SFT, DPO, LoRA, QLoRA, and RLHF; tracked experiments with MLflow and benchmarked quantized models for latency and cost/token.
- Streamlined OPRO-style evaluation with deepeval g-val, Grouse, RAGAS, and LLM-Judge, leveraging faithfulness, style, contextual relevance, and correctness metrics, accelerating iteration cycles by 3x.
- Containerized RAG with Docker & FastAPI, deployed on AWS (Bedrock/Lambda/sagemaker), automated CI/CD, integrated observability with Datadog, Prometheus, Grafana, and built real-time Kafka + Spark pipelines (5K events/sec), improving data freshness by 40% and search relevance by 22%.
Data Scientist ā Stylumia (Karnataka, India)
Sep 2020 ā Sep 2023
- Architected distributed training infrastructure on serverless Kubernetes (Knative) with Kubeflow, automated CD pipeline using GitHub Actions and Terraform, and integrated observability metrics using Prometheus and Grafana.
- Fine-Tuned Sentence Transformers with DeepSpeed ZeRO-3 and FlashAttention on Enterprise Data to generate contextual embeddings for user, product title & description, and visualized relationships using Neo4j graph database.
- Implemented Elasticsearch-based search engine with logic-driven ranking, weightage-based scoring, trend analysis, and category-based filtering to enhance search relevancy and user experience.
- Automated clothing taxonomy generation using GPT-4 via Chain-of-Thought (CoT) prompt techniques, improving product categorization efficiency and accuracy.
- Led statistical evaluation (A/B testing, hypothesis testing) of recommender systems(Recsys) and DBSCAN.
- Built attribute extraction using Visual Transformers & Detectron2, improving catalog search and future trends.
- Led annotation team for text NER and instance segmentation engine, ensuring high-quality labeled datasets delivery on-time.
Machine Learning Engineer Intern ā SoluLab (Gujarat, India)
Jan 2020 ā Jun 2020
- Revamped an end-to-end credit risk prediction system using PySpark (data processing), LightGBM (82% accuracy), and MLflow (experiment tracking), reducing default risk by 18% through automated weekly retraining.
- Deployed scalable ML pipelines with Model Serving, implemented A/B testing (improved F1-score by 15%), Feature Store integration, CDF-triggered retraining upon data drift detection.
- Automated CI/CD pipelines using Azure Databricks Asset Bundles (DAB), enabling 90% faster deployments with Infrastructure-as-Code, Auto-retraining when data drift exceeds 10%
Machine Learning Engineer Intern ā Career Launcher (Remote)
May 2019 ā Aug 2019
- Trained a multi-label object detection model using YOLOv8, achieving 0.7 mAP by tuning hyperparameters with Optuna, saving checkpoints, and validating performance on held-out data.
- Converted the optimized model from PyTorch ā ONNX ā TensorRT, delivering 30 FPS inference on an NVIDIA T4 GPU, ensuring deployment-ready performance.
- Managed dataset labeling of 24 object classes using CVAT, led the annotation team for on-time delivery, and built a real-time Power BI dashboard to monitor detection metrics.
Skills
- Programming & Databases: Python, PostgreSQL, PySpark, MongoDB, Pinecone, PgVector, Qdrant
- Libraries & Frameworks: PyTorch, TensorFlow, ONNX, Spacy, Gensim, timm, Scikit-learn, Hugging Face Transformers
- Development & Tools: FastAPI/Flask/Falcon, Streamlit, Docker/Podman, GitHub, LangChain, CrewAI
- Cloud & MLOps: AWS(ECR, EKS, Sagemaker, Lambda, Bedrock, Redshift), GCP(Vertex AI, GKE, Bigquery), Kubernetes, Kserve, Knative, Keda, Kubeflow, MLflow, ZenML, Comet, Datadog, Prometheus & Grafana
Side projects
In my free time, I like to work on side projects to keep my skill sharp and try out new tech. Here is a list of my current projects:
- MirrorMuse - Developed an chat platform that crawls resources to generate Q&A for interview preparation.
- LLM PROD - LLM PROD fine-tunes, deploys, serves LLMs efficiently.
- Price Predictor - Price Predictor leverages ML, NLP for price Predicting and mini projects.
- CreditGuard - Credit Default Predictor: Clean, Train, Track, Test, Serve, Analyze.
- AI-NLP-Lab - Advanced NLP projects: medical QA, NER, classification, deployment.
- SafeMedAI - Built a scalable and reproducible pipeline using openFDA datasets to detect and predict Adverse Drug Reactions, with integrated observability and distributed processing for large-scale data.
Hobbies
Outside of programming, I enjoy trekking, playing badminton, reading books, and occasionally going out. I believe having hobbies beyond coding is essential for maintaining good mental health.
Iām also deeply interested in self-improvement, nutrition, and positive psychology.
Social Media
Links to all my social accounts:
Business inquiries
If you want to work with me on a please contact me via email at aquibalicool3@gmail.com