Skip to content
View silsgah's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report silsgah

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. 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
silsgah/README.md

Silas Kwabla Gah

Applied AI & Software Architect
Agentic Systems · LLMs · RAG · AI Evaluation · Cloud Engineering

Portfolio · LinkedIn · Email


I design and build AI and software systems from technical discovery and proof-of-value through implementation, deployment, evaluation, and operational handover.

My work sits at the intersection of agentic AI, production LLM systems, enterprise software architecture, and reliable AI evaluation. I also conduct PhD research in computer vision and foundation-model adaptation.

Selected work

Tool use, live repository context, approval gates, persistent memory, context reduction, bounded sub-agents, vLLM/Transformers inference, QLoRA, FastAPI, Docker, and automated testing.

An eight-part engineering course covering agent loops, permissions, durable checkpoints, replay, Docker sandboxing, context management, memory, parallel sub-agents, evaluation, and issue-to-PR workflows.

Full-stack OCR/RAG system with Anthropic, OpenAI, and Ollama support; authentication; organization-aware application workflows; invoice extraction; streaming; deployment configuration; and an explicit governance and production-hardening model.

Agentic legal research system using RAG, LangGraph, DeepEval evaluation, LLM observability, FastAPI/Next.js, and automated legal-data ingestion.

End-to-end LLM/RAG engineering across ingestion, retrieval, query expansion, reranking, fine-tuning, inference, APIs, observability, deployment, and CI/CD.

PyTorch, Triton, and native CUDA/C++ implementations of deep-learning operators with correctness validation and latency benchmarking.


What I work on

  • Agentic AI: tool use, coding agents, sub-agent coordination, memory, context engineering, evaluation
  • LLM systems: RAG, vector retrieval, reranking, structured outputs, model/provider integration
  • Enterprise AI delivery: discovery, proof-of-value, architecture, deployment, handover, adoption
  • Software architecture: Python/FastAPI, C#/.NET, TypeScript/Next.js, REST/gRPC, SQL/PostgreSQL
  • Cloud & DevOps: AWS, GCP, Docker, Kubernetes, GitHub Actions, CI/CD
  • Research: foundation models, computer vision, 3D scene understanding, multimodal reasoning, calibration and robustness

Engineering stack

AI / ML
Python · PyTorch · Hugging Face · vLLM · Qwen · LangChain · LangGraph · LoRA/QLoRA · MLflow · Weights & Biases

Software
C# · .NET · ASP.NET · FastAPI · TypeScript · Next.js · REST · gRPC · SQL Server · PostgreSQL · Redis

Cloud / Platform
AWS · GCP · Docker · Kubernetes · GitHub Actions · CI/CD

Research

PhD candidate in Computer Science at the University of Ghana. My research focuses on robust visual perception under low supervision, inference-time coordination of frozen foundation-model priors, open-vocabulary segmentation, 3D scene understanding, and multimodal reasoning.


Open to Applied AI, AI Architecture, Agentic Systems, LLM Engineering, and senior software/AI delivery opportunities.

Pinned Loading

  1. gpu-kernel-lab gpu-kernel-lab Public

    Python

  2. rag-llm-system rag-llm-system Public

    RAG system for news portal across the country Ghana

    Python 1

  3. coding-agent-course coding-agent-course Public

    Building a Coding Agent From Scratch — 8-week course by EED Soft Consult

    Python 1

  4. coding_agent coding_agent Public

    Python

  5. Ghana_Legal_AI Ghana_Legal_AI Public

    Python

  6. ocr_llg_rag_responsible_AI_Governance ocr_llg_rag_responsible_AI_Governance Public

    Python