Bio

Ayesh is a founder and AI systems researcher building neuromorphic, brain-inspired models designed for strict power, latency, and reliability constraints. His work spans spiking models, hardware-aware learning, and full software stacks—compilers to runtimes—that translate research into deployable edge systems. Shaped by production deployments, he focuses on efficiency, robustness, and specialized models over ever-larger architectures.

Education & Industry

2025 -

I am a founder at CVision. Here we build AI-first automation systems that transform business operations. Currently focused on neuromorphic computing and brain-inspired AI for edge deployment under strict power and latency constraints.

2025 - 2025

I moved to pursuing a greater level of R&D in the energy-efficient AI systems industry by first making my way to the American industry as a consultant.

Serving as an AI consultant to multiple founders, I learnt a lot about the founder journey, business operations, and guiding them from discovery to delivery.

2024 - 2025

I was an ML, CV, and SW Design Engineer (meaning, practically everything) at Cowlar (YC W17). This was my first hands-on experience with real-world AI systems, hardware, and taking lead in research and development. I majorly contributed to late refinement stages in the viaPhoton Fiber Organizer and directed the R&D of the Al Meera Smart Shopping Cart (first in the Middle East).

Having to handle pretty much everything in a very fast-paced industrial environment, I learned a lot about the importance of efficiency, robustness, and scalability, and even deployed edge ML on Qualcomm's Hexagon NPU for ultra-fast, low-latency inference.

2021 - 2025

BESE at the National University of Sciences and Technology with a specialization in deep learning, and a heavy emphasis on mathematical and theoretical foundations. This is where I first got into deep learning and working with resource-constrained systems.

Blogs

I write about energy-efficient AI systems, the intersection of research and product, my founder journey, problems in the industry, life, and more. Here's some of my favorite posts:

Dec 2025 Why Neuromorphic Computing Matters for Edge AI
May 2025 The Unfair Advantage (and Issue) with Pakistan's Software Industry
Dec 2024 The Gap Between Research and the Market

Pet Projects

Industrial projects are not included here. This list is some of my favorite pet projects. See my up-to-date projects on my GitHub.

Vaultbot — An intelligent customer service assistant for NUST's on-campus banks, built using open-source LLMs and platforms. Developed as per request of Prof. Dr. Faisal Shafait at SEECS, NUST.

M.A.S.S. — Massively available surveillance system created using a network of IoT-cams and activity recognition to classify and track malicious activity across the connected areas.

GPT-2 from Scratch — Recreating GPT-2's decoder-only transformer architecture and training it on FineWeb-Edu to predict the next token in a sequence.

Publications

Neuromorphic Intermediate Representation for Type 1 Compute FPGA Prototype
Manuscript in preparation; internal review phase (2025)
Faras Siddiqui, Ayesh Ahmad