LlamasHub
Deploy and manage multiple self-hosted language models through one OpenAI-compatible API.
Explore source ↗Computer Vision · Generative AI · Research
Calculating experience…
Engineering intelligence,
from pixels to systems.
I'm Soumen Sardar, a software engineer and published researcher working across computer vision, deep learning, embedded applications, developer tools and AI infrastructure. I enjoy taking ideas from algorithms to working systems.
Selected work
Featured engineering work
A cross-section of AI systems, security imaging, practical computer vision and deployed-system thinking.
cpp_fish_cv
Real-world fish counting and centroid-based tracking with C++ and OpenCV. Public GitHub repository.
ContentCreator
AI-powered content creation project, exploring generative workflows and practical applications.
From the engineering desk
Engineering updates
Selected posts and project notes shared on LinkedIn.
LinkedIn update 1
View on LinkedIn ↗LinkedIn update 2
View on LinkedIn ↗LinkedIn update 3
View on LinkedIn ↗Portfolio / repository index
Projects worth exploring
25 projects, grouped by technology and problem domain. Public repositories link directly to source.
Peer-reviewed chapters & research preprint
Research & Publications
Published research spanning computer vision, medical AI, security imaging and deterministic clutter measurement.
Researcher identifier
View ORCID profile ↗ORCID iD: 0009-0003-8236-4923
SNOCS: Deterministic Scale-Normalized Occupancy Clutter Score
A deterministic, scale-normalized approach to quantifying spatial occupancy clutter.
Read publication ↗AI in Cancer Screening and Early Detection
Research on applications of artificial intelligence to cancer screening and early detection.
Read publication ↗Application of Deep Learning Algorithms to Terahertz Images for Detection of Concealed Objects
Deep learning for detecting concealed objects in terahertz security-screening imagery.
Read publication ↗A Study to Find Affordable AI Techniques for Early Parkinson's Disease Detection
Exploration of accessible AI approaches for early Parkinson's disease detection.
Read publication ↗Technical essays · Medium
Featured writing
Notes on the mathematics and engineering behind machine learning. Selected articles from my Medium portfolio.
Oversimplified MLOps & DLOps — Part 2
From experiments to reproducible models: experiment tracking, model versioning and model registration for real-world ML engineering.
Read article ↗My LLM — Was It Safe? Adding Guardrails, Langfuse, and Persistent Storage
Building safer, observable LLM systems with guardrails, Langfuse tracing, and persistent storage.
Read on Medium ↗Statistics for Generative AI — KL Divergence
An intuitive walk-through of probability distributions and why KL divergence matters for generative models.
Read on Medium ↗More practical writing on PyTorch, GPU memory, reproducibility and Python concurrency.
Explore all Medium articles ↗About my work
Research mindset. Engineering discipline.
My interests bridge learning-based algorithms and dependable software: vision pipelines, multi-object tracking, AI model serving, system integration and low-level development. Alongside product engineering, I have published research related to imaging and AI.
Beyond the code
Find me around the web
Technical writing, professional experience, coding challenges and developer-community contributions.
Connect with me for computer vision, Applied AI, research and engineering opportunities.
Connect on LinkedIn ↗Medium
Explore my technical writing and ideas beyond the repositories.
Read on Medium ↗HackerRank
Explore my coding and algorithm practice profile.
View HackerRank ↗Stack Overflow
Visit my Stack Overflow developer profile.
View Stack Overflow ↗Connect
Let's build something useful.
Explore the code, discuss a project or connect around applied AI and computer vision.