AI/ML Engineer & Full-Stack Builder
View RésuméOpen to remote & freelance ISB
I build AI systems and ship them — computer-vision pipelines, LLM-powered tooling, and full-stack apps that actually reach production. Electronics student at Quaid-i-Azam University, going deep on machine learning while building real products.
I architect the systems; AI agents do the heavy lifting. Build fast, ship publicly, learn through things that actually work. If you need someone to take an ML idea from prototype to deployed product — that’s the whole game.
AI-Powered Churn Retention · SaaS
B2B SaaS platform for spotting churn risk before revenue disappears. Combines customer behavior, account health, support pressure, billing signals, and AI-assisted recommendations into a single operating surface for retention teams, operators, founders, and customer-success leaders.
Generative Web Tool · Monorepo
Nostalgia-heavy web tool that converts public websites into retro-themed HTML artifacts — Windows XP, Windows 98, GeoCities, CRT terminal, Mac Classic, Netscape, ANSI/BBS, and early Blogspot. Each era gets its own chrome, spacing, and era-correct weird confidence.
LLM Tooling · Shopify
AI-powered Shopify store builder. Input a niche, choose an audience, and watch AI generate themes, copy, email flows, ad creatives, and policy pages — deploying a fully configured live Shopify store in under 60 seconds.
Finding real bugs in production-grade AI infrastructure and fixing them at the source.
Fixed Amazon Bedrock toolResult co-location across model families — preventing validation failures during agent conversations.
Resolved SQL parser edge case — parenthesized multi-expression queries were incorrectly interpreted as subqueries.
Fixed tick-label formatting for small numbers in exponential notation — correct axis rendering across chart types.
Honored max_det in RT-DETR decoder exports and validation — consistent detection limits across training and deployment.
Fixed resampling engine for BusinessDay and CustomBusinessDay with closed="right" — improving time-series bin alignment.
Extended drop_params support for embedding calls across Azure and OpenAI-compatible providers — consistent cross-provider behavior.
Fixed proxy auth logic — bypassed budget validation for model discovery endpoints, restoring internal API access without weakening security.
Preserved inline nested objects in JSON Schema → Pydantic conversion instead of collapsing to generic dicts — better schema fidelity.
Auto-converts Python dicts and lists to JSON before passing to LLMs — improves interoperability with structured-output pipelines.
From model weights to a container behind a load balancer.
I’m an electronics student at Quaid-i-Azam University (Class of ’27) going deep on machine learning while building real AI products. My edge is hardware-aware thinking: signal processing, quantization, and edge inference come naturally because the electronics foundation is already there.
I find bugs in production-grade AI infrastructure and fix them at the source. Electronics background is a research differentiator — a natural edge in hardware-aware ML, quantization, and FPGA inference.
Deep learning research at NCP Islamabad. Working on machine learning and deep learning systems, applying modern architectures to real research problems.
DSP in a live production environment: real-time audio pipelines and broadcast transmission. Applied signal-processing theory at production scale.
Focus on transformers, attention mechanisms, clustering, computer vision, and edge-AI deployment.
Foundations in computing, mathematics, and structured problem solving.
Live contribution graph, stats, and recent activity from my GitHub.
Loading contribution graph…
Loading live data…
Open to remote roles, freelance, and interesting ML problems. The fastest line is email.
hasnaat.hussain.2@gmail.com