Agentic AI applications
Designs agent workflows with AWS Bedrock, Strands, LangChain, LangGraph, tool registries, MCP context management, multi-agent coordination, and persistent or session-based memory architectures.
Senior AI & Machine Learning Engineer
AI and machine learning engineer with professional work spanning production conversational systems, retrieval pipelines, model fine-tuning, and cloud deployment. He has delivered AWS Bedrock agents, Azure-hosted assistants, LangGraph workflows, and domain-specific LLM applications. His experience also includes data science, computer vision, forecasting, and engineering team leadership.
Designs agent workflows with AWS Bedrock, Strands, LangChain, LangGraph, tool registries, MCP context management, multi-agent coordination, and persistent or session-based memory architectures.
Builds domain-specific conversational systems using retrieval augmentation, metadata filtering, prompt engineering, dataset curation, streaming inference, and parameter-efficient fine-tuning with distributed GPU training.
Develops classification, forecasting, NLP, and computer vision models with Python frameworks, supported by feature extraction, evaluation, dataset analysis, dashboards, and production data processing.
Deploys AI applications on AWS and Azure, implementing asynchronous backend infrastructure, Lambda orchestration, FastAPI or Flask services, and collaborative engineering workflows for production systems.
Deploys enterprise agents on AWS Bedrock and AgentCore, implementing managed execution, persistent and session memory, Strands orchestration, tool registries, retrieval pipelines, and serverless integration for production workloads.
Built production AI applications on Azure, including loyalty segmentation and concierge assistants, while designing MCP context management, modular LLM architecture, and LangChain or LangGraph agent-driven enterprise workflows.
Developed production conversational agents with streaming inference and retrieval, led NLP and data science teams, and managed domain-specific datasets, evaluations, backend architecture, and parameter-efficient LLM adaptation workflows.
Developed stock forecasting models and embedded technical indicators, including news-derived signals, while testing alternative time windows and forecast horizons for the time-series work undertaken during this contract engagement.
Built a deep learning classification pipeline for human behavior analysis, extracting facial features and connecting feature engineering with model development across the end-to-end work delivered during this consulting contract.
Delivered client machine learning, NLP, deep learning, and analytical dashboard projects, working with structured and unstructured datasets and connecting data sources with automated visualizations and predictive modeling workflows.
Developed item classification and action recognition models, evaluated training and model performance, integrated real-time processing into a production pipeline, and delivered data analysis and visualization using Tableau.
Developed sentiment analysis using scraped Twitter data, applying NLP to opinions and predictions while using Python APIs to automate repetitive reports and reduce manual data processing effort.
Highlights from professional projects, product delivery, and specialist work.
Implemented production agents using AWS Bedrock and AgentCore managed execution, with dual-memory design and Strands orchestration. Built custom tool registries and metadata-filtered retrieval using S3 Vectors, connecting Lambda functions for dynamic query processing and serverless workflows.
Built Azure-hosted GPT assistants for concierge interactions and campaign management. Combined customer segmentation, vector-based semantic search, multimodal interaction, MCP context management, and LangGraph workflows to support contextual assistance and automated enterprise tasks.
Developed conversational agents with streaming inference, retrieval augmentation, domain dataset curation, and evaluation workflows. Adapted language models using LoRA and QLoRA with DeepSpeed distributed training, and implemented asynchronous backend infrastructure for production interaction.
Bahria University, Islamabad
Feb 2016 – Jan 2020
FAST NUCES, Islamabad
Feb 2022 – Jan 2024
Dice Analytics
Builds retrieval, memory, tool use, and orchestration into deployed assistants, connecting model architecture with asynchronous APIs, cloud services, evaluation workflows, and engineering team delivery.
Combines curated domain datasets, parameter-efficient fine-tuning, distributed GPU training, and evaluation protocols to adapt conversational models and support reliable real-time inference in production applications.
Share your requirements, current challenges, and delivery goals. Botmer will review the brief and help arrange the next step.