Skip to content
View shahriar-ahmed-seam's full-sized avatar

Block or report shahriar-ahmed-seam

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse

Shahriar Ahmed Seam

Open to work  Focus


About

I'm an AI/ML Engineer and final-year CSE student at BUET. I build intelligent systems and take them from research to production — LLM and agentic applications, applied computer vision, and the full-stack engineering that ships them.

First-author on SYNAPSE-Net, a framework for robust brain lesion segmentation. Currently focused on agentic AI, RAG systems, and generative models.


Experience

Founder & AI/ML Engineer · Somokolon Labs   Jan 2026 – Present

Founded and run an AI and software studio that ships LLM, RAG, generative-AI, and full-stack products end to end — including production software for small businesses, from requirements through deployment. Own the full lifecycle: research, engineering, CI/CD, cloud deployment, and monitoring across the studio's shared infrastructure.


What I Build

LLM & Agentic Systems

Multi-agent orchestration (plan → research → code → critique), RAG pipelines over documents, and offline-capable agents with swappable local/cloud LLM providers.

Applied ML, CV & Systems

Generative & diffusion models, on-device inference (ONNX, edge AI), and performance-critical systems — including a vector database built from scratch with AVX-512 kernels.

Production Engineering

Fault-tolerant microservices, offline-first architectures for low-connectivity environments, and full-stack products taken from research notebook to deployed, monitored infrastructure.


Featured Projects

A selected few from 85+ repositories. Browse all →

Secure, ephemeral code-execution infrastructure for AI agents — isolated Docker sandboxes behind a FastAPI gateway.

Docker · FastAPI · Next.js · Go

Live Repo

Multi-character AI chat across web, Android, and Telegram — one FastAPI backend with persistent memory and RAG.

FastAPI · RAG · React · PostgreSQL

Repo

Private, offline-first AI assistant for Windows 11 powered by local LLMs, with a native WinUI 3 desktop app.

C# · WinUI 3 · Next.js · Ollama

Repo

High-throughput notification engine — a million messages in five minutes over RabbitMQ and Redis.

FastAPI · RabbitMQ · Redis · Next.js

Live Repo

Full-stack, AI-native ERP (HR, CRM, Inventory, Finance) with private on-device AI. Runs standalone for demos.

Next.js · Prisma · Ollama · TypeScript

Live Repo

SYNAPSE-Net  Research

A framework with lesion-aware hierarchical gating for robust brain lesion segmentation. First-author publication.

PyTorch · Computer Vision · Medical Imaging

Paper


Tech Stack

Languages Languages
AI / ML AI/ML HuggingFace LangChain LlamaIndex Ollama
Backend Backend RabbitMQ
Frontend Frontend
Data / Infra Infra Pinecone Chroma
Cloud Cloud

Certifications

Issuer Certification
Microsoft AI & ML Engineering — end-to-end ML infrastructure on Azure
AWS Solutions Architect
IBM Generative AI with LLMs  ·  RAG & Agentic AI  ·  Multimodal Generative AI  ·  DevOps, Cloud & Agile
CNCF Certified Kubernetes Administrator (CKA)
Google Cloud Generative AI
DeepLearning.AI Deep Neural Networks — Hyperparameter Tuning & Optimization

All certifications


GitHub Stats

GitHub Stats Top Languages

Stats & language cards are self-hosted — built by me 😉

Contribution Graph

Competitive Programming & Kaggle


footer

Pinned Loading

  1. AI-Code-Sandbox AI-Code-Sandbox Public

    Secure, ephemeral code-execution infrastructure for AI agents - isolated Docker sandboxes, strict resource limits, FastAPI gateway, and a v0-style Next.js playground.

    TypeScript 1

  2. Character-Chat-AI Character-Chat-AI Public

    Multi-character AI chat across web (PWA), Android, and Telegram - one FastAPI backend with persistent memory, auth, and swappable LLM providers (OpenRouter/Ollama/Gemini).

    Python 1

  3. Hyper-Match-Engine Hyper-Match-Engine Public

    Low-latency limit order matching engine: deterministic, zero-hot-path-allocation C++ matching core, a Rust HTTP/WebSocket gateway, a binary wire protocol, and a real-time web console.

    C++ 1

  4. gravitas gravitas Public

    A cinematic, physically-accurate 3D gravitational N-body simulator in the browser. RK4 physics, tidal disruption, accretion disks, Lagrange points and 15 preset scenarios. Built with React, Three.j…

    JavaScript 1

  5. Vector-Vault-DB Vector-Vault-DB Public

    A high-performance vector database built from scratch in C++17 with Python bindings: HNSW/IVF ANN indexes, AVX-512 distance kernels, a custom arena allocator, and a memory-mapped snapshot format.

    C++ 1

  6. p2p-portal-drop p2p-portal-drop Public

    WrapDrive — LAN-only P2P file sharing: browser portals pair with native apps via QR/pairing-code and transfer over WebRTC. Home of the p2p-portal-drop web SDK + signaling server. Built on LocalSend.

    Dart 1