Looking for a job

Syed Muhammad Sarim Ahsan

@syedmuhammadsarimahs

Message

I build open-source LLMs, CUDA kernels, and production AI systems from the ground up.

Pakistan

Message

What I'm looking for

I'm looking to build and understand AI systems from the ground up, contributing to LLM research, interpretability, efficient inference, model training, CUDA optimization, and reliable production AI infrastructure.

I build open-source LLMs, interpretability tools, and reproducible AI systems, from decoder-only Transformers and CUDA kernels to RAG platforms. At Global Integration System, I engineer Portalersm, a production student recruitment platform serving 600+ active users across the UK.

My research examines Transformer architecture, hallucination detection, efficient inference, and cross-environment robustness. I’ve built and evaluated systems including InterpLens, Nova-MoE, Nova-141M, a pure NumPy GPT-2 inference engine, speculative decoding benchmarks, and preference-optimization experiments that reduced toxic outputs by 84.2%.

I enjoy understanding AI systems from the ground up—training models, profiling GPU kernels, and turning research ideas into usable tools and production infrastructure.

Experience

Work history, roles, and key accomplishments

Current

Software Engineer

Current

Global Integration System (GIS)

Jan 2025 - Present (1 year 8 months)

Engineered and maintain Portalersm, a production student recruitment platform serving 600+ active users across the UK. Architected scalable full-stack features, Node.js backend services, secure role-based access control (RBAC), Docker containerization, and zero-downtime CI/CD deployment pipelines using GitHub Actions.

Node Docker GitHub Actions RBAC CI CD Full Stack