Skip to content
Botmer International®
Botmer talent profile

Ahsan Sarwar

Senior AI & Computer Vision Engineer

AI engineer with eight years of experience specializing in computer vision, NVIDIA inference tooling, and edge deployment. He develops detection, segmentation, tracking, and re-identification systems using DeepStream, Triton, TensorRT, and Jetson hardware. His work also covers document processing, retrieval-based assistants, recommendation pipelines, and technical team leadership.

8+ years
AI engineering experience
NVIDIA AI stack
real-time vision deployment
Vision & LLMs
applied AI specialization

Islamabad, Pakistan

Meet the engineer

AI engineering with production depth

Where Ahsan adds value

  • Designed DeepStream and Triton ensemble pipelines for real-time inference across A100 and Jetson AGX hardware with multiple vision models.
  • Developed a custom re-identification model that improved FairMOT tracking accuracy, alongside GPU-based preprocessing and postprocessing within inference pipelines.
  • Optimized detection, segmentation, pose estimation, and tracking models using ONNX and TensorRT while integrating Redis, Azure storage, and SQL.
  • Built LLM document deduplication, extraction, splitting, and title-assignment workflows using prompt engineering and custom logic for enterprise processing.
Explore professional experience
Core competencies

AI systems grounded in software engineering

01

Computer vision systems

Develops object detection, classification, semantic segmentation, tracking, pose estimation, and re-identification models, supported by custom datasets, training workflows, evaluation, and production backend integration.

  • PyTorch
  • TensorFlow
  • OpenCV
  • DETR
  • AlphaPose
  • FairMOT
02

NVIDIA inference optimization

Builds real-time multi-model pipelines with DeepStream, Triton, GStreamer, ONNX, and TensorRT, including GPU preprocessing and postprocessing for datacenter and Jetson edge hardware.

  • DeepStream
  • Triton
  • TensorRT
  • ONNX
  • GStreamer
  • Jetson
03

LLM application engineering

Builds document extraction, duplicate detection, automated titling, recommendation pipelines, and retrieval-based chatbots using prompt-engineered LLMs, domain logic, and modular enterprise processing workflows.

  • LLMs
  • RAG
  • LangChain
  • Python
  • Prompt engineering
04

Production ML delivery

Packages inference workflows with Docker, orchestrates models through Apache Airflow, and integrates Redis, Azure Blob Storage, and SQL while maintaining documentation and code review practices.

  • Docker
  • Apache Airflow
  • Redis
  • Azure Blob Storage
  • SQL
  • Python
Experience

From models to product delivery

Mar 2024 – PresentInfixsystems Pakistan

Senior AI Consultant

Delivers applied AI systems spanning travel recommendations, workplace PPE detection, printed-material defect detection, safety analytics, and retrieval-based assistance, combining computer vision pipelines with prompt-engineered language model applications.

  • Built review-based travel recommendation pipelines.
  • Implemented PPE and printed-material defect detection systems.
  • Developed a RAG-based support chatbot.
Feb 2024 – PresentDeepLearningAnalytics Canada

AI Engineer

Builds modular document processing pipelines that combine language models and custom logic to detect duplicates, extract information, split large files, and assign titles across varied enterprise document structures.

  • Implemented LLM-based duplicate document detection.
  • Built extraction workflows for complex documents.
  • Automated document splitting and accurate title assignment.
Jun 2024 – Dec 2024Nexta Saudi Arabia

AI Engineer

Developed CCTV-based coffee shop analytics, measuring cup production, espresso waste, and employee movement by zone, with NVIDIA inference optimization to support real-time operational performance insights and reporting.

  • Built zone-based staff movement and production monitoring.
  • Optimized video analytics with DeepStream and TensorRT.
Sep 2023 – Feb 2024Morrow AI Saudi Arabia

AI Engineer

Developed retail CCTV analytics for visitor entry, exit, and dwell time, using NVIDIA tooling and edge pipelines to process footage and aggregate behavior data for visualization dashboards.

  • Implemented customer entry and exit monitoring.
  • Built retail zone dwell-time analytics using DeepStream and TensorRT.
Oct 2021 – Sep 2024Dlision Pakistan

Team Lead

Led real-time vision projects and model deployment, designing ensemble inference pipelines, custom re-identification, GPU processing, orchestration, and technical documentation across datacenter and Jetson deployment environments for client applications.

  • Improved FairMOT MOTA from 0.47 to 0.91 using custom re-identification.
  • Built DeepStream and Triton ensembles for A100 and Jetson AGX.
  • Managed Dockerized inference pipelines with Apache Airflow.
Feb 2019 – Oct 2021Dlision Pakistan

Machine Learning Developer

Developed vision models and datasets for detection, classification, segmentation, tracking, pose estimation, and number recognition, optimizing inference and integrating deployed systems with data storage and backend services.

  • Optimized inference using ONNX and TensorRT.
  • Integrated models with Redis, Azure Blob Storage, and SQL.
  • Created custom datasets and evaluation workflows.
Feb 2018 – Jul 2018HITEC University Incubation Center

Research Assistant

Applied classical image processing to garment defect detection and built beverage brand classification models, contributing to machine learning research initiatives and a funded research grant at the university incubation center.

  • Developed garment defect image processing.
  • Built classification for beverage brands.
  • Secured a PKR 0.5 million research grant.
Selected work

AI systems built for practical use

Highlights from professional projects, product delivery, and specialist work.

Selected project

Real-Time Ensemble Inference

Designed a DeepStream and Triton ensemble pipeline integrating DETR, FCHarDNet, and AlphaPose for real-time vision workloads. Deployed across A100 and Jetson AGX hardware, combining multi-model orchestration with optimized inference and documented production delivery workflows.

  • DeepStream
  • Triton
  • DETR
  • Jetson
Selected project

Re-ID Model Optimization

Developed custom re-identification to improve FairMOT tracking and integrated GPU-accelerated preprocessing and postprocessing into DeepStream pipelines. Packaged inference in Dockerized Triton environments and used Apache Airflow to manage machine learning deployment workflows.

  • FairMOT
  • TensorRT
  • Docker
  • Apache Airflow
Selected project

Duplicate Document Detection

Built an LLM-based document pipeline combining prompt engineering and custom logic to identify duplicate files, extract required information, split large documents, and automatically assign descriptive titles. Designed modular processing for varied enterprise document structures and formats.

  • LLMs
  • Python
  • Prompt engineering
Education & professional focus

A foundation for practical delivery

Education & qualifications

MS in Computer Science

HITEC University, Taxila

Feb 2016 – Jan 2019

BS in Computer Science

UET, Peshawar

Sep 2011 – Jun 2015

Upwork Skill Certification – Machine Learning

Upwork

October 2024

Building Video AI Applications at the Edge on Jetson Nano

NVIDIA DLI

July 2023

Professional focusReal-time vision deployment optimization

Develops inference pipelines across NVIDIA datacenter and edge hardware, combining model optimization, GPU processing, GStreamer integration, and production orchestration for practical computer vision applications.

Professional focusApplied AI workflow automation

Builds language model and vision solutions for document processing, recommendations, safety inspection, and video analytics, supported by domain-specific logic, custom datasets, and modular deployment workflows.

Talk to Botmer

See if Ahsan fits your AI engineering team

Share your requirements, current challenges, and delivery goals. Botmer will review the brief and help arrange the next step.

Request a talent match