Agentic AI and LLM systems
Develops multi-agent applications, prompts, LLM extraction, and natural-language database interfaces, integrating AWS Bedrock, Lambda, S3, and language-model tooling into practical operational workflows for business users.
Senior Data Scientist & AI Engineer
Data scientist and AI engineer experienced in agentic applications, NLP, phishing detection, computer vision, and machine-learning delivery. Works with Python, AWS Bedrock, PyTorch, TensorFlow, Spark, and MLflow. His background includes SOC data transformation, BERT email security, natural-language database access, transcription, sentiment analysis, and sports tracking.
Lahore, Pakistan
Develops multi-agent applications, prompts, LLM extraction, and natural-language database interfaces, integrating AWS Bedrock, Lambda, S3, and language-model tooling into practical operational workflows for business users.
Trains transformer models for phishing and scam detection using email text, metadata, OCR, and document extraction, while applying precision optimization and continual-learning research to email security.
Builds image, video, and audio analysis using CNNs, object detection, speech models, denoising, transcription, and OCR for engagement scoring, sports tracking, and structured multimedia interpretation.
Uses Python, SQL, Spark, Databricks, Pandas, and NumPy for data processing, with MLflow, Docker, Flask, SageMaker, and vector databases supporting model development and deployment workflows.
Develops an agentic AI system for SOC analysts to analyze, parse, and transform security data. Uses AWS Bedrock multi-agent orchestration, Lambda, and S3 to coordinate agent workflows and file operations.
Contributed end-to-end phishing and scam detection using textual signals and metadata. Designed and trained BERT models for PDF attachments with OCR and text extraction, then optimized precision against large volumes of customer email traffic.
Built LLM and analytics tools for business and customer-support operations. Developed GPT-4 natural-language-to-SQL access, fine-tuned information extraction, and applied review sentiment analysis and clustering to improve issue understanding and resolution workflows.
Developed audio transcription and computer-vision systems spanning virtual engagement and tennis analytics. Combined Hugging Face speech models, denoising, grammar correction, CNN-based emotion detection, YOLOv5 tracking, and OCR-based score interpretation in multimedia workflows.
Highlights from professional projects, product delivery, and specialist work.
Developed a security-operations workflow using AWS Bedrock with one orchestrator and four collaborator agents. Integrated Lambda and S3 file operations to analyze, parse, and transform security data, reducing transformation time by 80%.
Designed and trained BERT models that use OCR and extracted attachment text to identify phishing and scams. Combined textual and metadata features and optimized detection precision against millions of emails in live customer environments.
Built a GPT-4 and LangChain text-to-SQL system for business teams to query databases in natural language. Complemented database access with LLM email information extraction, app-review sentiment analysis, and K-means clustering for support workflows.
Built transcription with Hugging Face models, denoising, and grammar correction, alongside CNN-based facial-emotion detection for virtual meetings. Developed YOLOv5 tennis tracking with OCR score analysis and player profiling for structured sports intelligence.
Information Technology University, Lahore
2019 – 2021
GIFT University, Gujranwala
2015 – 2019
His recent work applies language models, multi-agent orchestration, document understanding, and phishing detection to security data and operational analyst workflows.
His experience spans NLP, computer vision, speech, and business analytics, connecting trained models with practical interfaces and domain-specific data processing.
Share your requirements, current challenges, and delivery goals. Botmer will review the brief and help arrange the next step.