AI/ML Engineer & Full-Stack Engineer with 5+ years designing and deploying production-grade AI systems, agentic workflows, and LLM-powered applications. Currently building AI agents for production banking at Ally Financial, after shipping products across HR tech, healthcare, and real estate.
name: "Hassan Naeem"
role: "AI/ML Engineer · Full-Stack Engineer"
location: "Michigan, USA 🇺🇸 (open to relocation)"
experience: 5+ years
education:
- M.S. Computer Science · Concordia University Saint Paul
- B.S. Computer Science · Sir Syed University of Eng. & Tech.
focus:
- Production LLM & agentic systems
- RAG pipelines & LLM orchestration
- Uncertainty-aware & calibrated ML
- Full-stack web & React Native mobile
verticals: [Banking, HR Tech, Healthcare, Real Estate, E-Commerce]
currently:
role: "Software Engineer @ Ally Financial (Contract)"
building: "Fresh Leads (live SaaS) · calibrated affect models · LLM agents"
open_to: "AI/ML & Full-Stack roles — relocation OK"
ask_me_about: [LLMs, Agents, RAG, React Native, Python, TypeScript]I'm an AI/ML Engineer and Full-Stack Engineer with 5+ years of experience designing and deploying production-grade AI systems, agentic workflows, and LLM-powered applications at scale.
I've translated complex AI research into high-impact commercial products across HR tech, healthcare, real estate, and e-commerce, integrating large language models, machine learning pipelines, and real-time data architectures with React, Next.js, React Native, Python, and cloud infrastructure on AWS and GCP.
Currently open to AI/ML and Full-Stack roles, full-time, contract, or freelance. Open to relocation.
Building AI agents and LLM-based automation systems for enterprise platforms, with intelligent workflows that streamline internal engineering and operations. Developing backend services and agent orchestration with Python, FastAPI, and AWS, integrating LLM capabilities into production banking infrastructure with a focus on reliability and compliance. Partnering with platform engineering teams on system design and the rollout of AI-assisted tooling across the organization.
Architected and deployed a production AI Agent system integrating VAPI.ai LLM, HeyGen, AWS Transcribe, and OpenAI APIs, enabling voice-command interactions, autonomous content generation, and real-time dashboard control. Designed agentic recruitment workflows with LLM-driven candidate ranking, interview-scheduling automation, and offer-letter generation via BambooHR integration. Built a GraphQL campaign-intelligence module with ML-driven audience segmentation and shipped CI/CD on AWS with Docker + Kubernetes.
Led frontend of Take Charge EV-charging platform (React Native, TypeScript, Django, Firebase) with real-time driver tracking, Authorize.net payments, and 100MS audio calling. Built DateJar matchmaking app (React Native, GraphQL, Agora SDK) with audio/video calling and in-app purchases. Optimized Nouvoprix marketplace with advanced product search and real-time messaging.
Developed GroceryDone4u e-commerce platform (React Native, Node.js, TypeScript, Firebase, Stripe, Google Maps) with multi-store inventory, food ordering, and real-time delivery tracking. Optimized Pacsend — a peer-to-peer package delivery platform (Angular, Node.js, GraphQL, Stripe) with real-time tracking and payment infrastructure.
A live SaaS that finds local businesses, verifies their phone numbers and emails at search time, confirms they're still trading, and scores each one against a buyer playbook, so two sellers on the same street get different lists. Public API, Stripe billing, required 2FA, and 571 passing tests.
Research on valence–arousal estimation that knows when it doesn't know. Deep evidential regression and split conformal prediction on AffectNet (3 seeds): CCC 0.692, ECE 0.011. Ships an in-browser ONNX demo that abstains when unsure, and no frames leave the device.
A neural network and autograd engine in
~200 lines of pure Python with zero libraries. A
Value graph runs backpropagation
through the chain rule, with Neuron → Layer → MLP
on top, and learns XOR. It ships with runnable lessons,
gradient checks, and an interactive visual
guide for stepping through backprop and watching
training happen live.
An advanced LLM intelligence layer combining RAG, multi-hop reasoning over knowledge graphs, continuous learning from user feedback, and parameter-efficient fine-tuning with LoRA / QLoRA. Multi-service architecture with a FastAPI backend and React frontend.
A Python agent that drives its own Chromium browser — no DOM shortcuts, no hidden state. Navigates nested iframes, does humanlike drag-and-drop on an HTML5 canvas, and reads the game board by sampling pixels. The same pattern generalizes to any canvas-based web app.
Production FastAPI service powering inbound carrier sales automation — load search, FMCSA MC-number verification, negotiation logging, and a webhook for the HappyRobot platform. Deployed on Fly.io with Docker + supervised processes and health-checked endpoints.
An interactive, gamified learning platform where beginners master programming through tutorials, quiz games, and hands-on coding sandboxes. XP, knowledge tree, achievements, and certificates across 43+ programming languages.
Architected an AI-driven property recommendation engine with Vertex AI and OpenAI for personalized user experiences. Built a hybrid React Native mobile app supporting immersive 3D property exploration and AR/VR hologram functionality for virtual furniture placement.
HIPAA-compliant healthcare platform with secure real-time patient-doctor communication via WebSockets and Firebase. Integrated Vertex AI and OpenAI to build an automated prescription-suggestion engine and predictive hospital-resource analytics.
AI-powered body measurement platform for e-commerce. Uses MediaPipe pose detection to extract 33 body landmarks from a photo, converts them to real-world measurements, and returns size recommendations against per-brand size charts.
Walking through the Python agent I shipped — nested iframe traversal, humanlike mouse drags, and reading a game board by sampling canvas pixels. The same pattern generalizes to Figma, Miro, and any legacy canvas-based internal tool.
A framework-free RAG you can actually read end-to-end — FAISS index, sentence-transformer embeddings, top-k retrieval, and an LLM call. No LangChain magic, just the four moving parts.
What an AI agent actually does when it "uses a tool" — the loop in ~100 lines. No framework, no hidden orchestration. Send tools, receive tool_use, run locally, send results, repeat.
Six chunking strategies — fixed, recursive, semantic, structural, parent-child, and contextual — with when to use each and the red flags to watch for. Everything I've learned building retrieval systems.
Twelve practical patterns for shipping with LLMs — role + task + constraints, few-shot that works, structured XML output, self-check pass, refuse-and-explain, and the anti-patterns to drop.
Score LLM generations against a reference with a strong model acting as judge. Rubric-based, returns structured JSON, aggregates correctness / completeness / clarity across a sample set.
The patterns I keep reusing — nested iframe traversal, global coordinate math, humanlike mouse drags, canvas screenshotting, and first-run browser install that self-heals. Starter kit for browser agents.
Concordia University Saint Paul
Sir Syed University of Engineering & Technology
I'm open to AI Engineer, Full-Stack, and Mobile opportunities, full-time, contract, or freelance. If you're shipping something interesting in GenAI, agentic systems, or product engineering, I'd love to hear from you.