Hi, I'm Naeem — I Build AI SystemsThat Run in Production
From fine-tuning AI models to building voice agents and multi-tenant RAG — I take AI from research to production, backed by full-stack engineering in Python, FastAPI, React, and Next.js.
Products Shipped
Live AI in production
RAG Accuracy
Multi-tenant, per-business isolation
Satisfaction Rate
Repeat clients & long-term work
From LLM research to production — I build AI that actually ships.
WhyWork WithMe
I don't just prototype models — I ship AI that runs in production, handles real users, and holds up under load.
Production, Not Prototypes
I've shipped AI across 4+ healthcare products and multiple SaaS platforms — fracture detection, voice agents, and multi-tenant RAG, all live and serving real users.
LLMs & Agentic AI
Multi-agent systems, agentic & hybrid RAG, and MCP tooling built with LangChain and LangGraph.
Voice AI, End to End
STT, TTS, and self-hosted voice agents — benchmarked, tuned, and deployed on vLLM with concurrent calls.
Self-Hosted Inference
I deploy models on Azure, Hugging Face, and self-hosted GPUs with vLLM — cutting cost and latency instead of just calling an API.
Full-Stack, Not Just Models
Python and FastAPI on the backend, React and Next.js on the front — I build the whole system around the AI, not just the model.
Measured by Results
71%→82% fracture accuracy, 94% RAG accuracy, 0.8s latency — I optimize for numbers that matter.
WhatI Do
From LLM agents and multi-tenant RAG to self-hosted voice agents and the full-stack apps around them — I build production AI end to end, research to deployment.
LLM & Agentic AI Systems
Multi-agent LLM systems, agentic and hybrid RAG, and MCP tooling — built with LangChain and LangGraph. Multi-tenant retrieval with per-business isolation, achieving 94% accuracy in production.
Voice AI
Self-hosted voice agents that take real calls — speech-to-text, text-to-speech, and voice cloning tuned for natural, multilingual conversation. Benchmarked and deployed on vLLM with concurrent-call support.
Full-Stack AI Apps
The complete system around the AI — FastAPI backends and React / Next.js frontends, from prototype to production. Scalable, maintainable architecture built to ship and grow.
SELECTED WORK
ThingsI've Built
Production AI systems across healthcare, real estate, and e-commerce — from research benchmarks to live deployments.
Clinical Speech UnderstandingNLP Pipeline & EMR Automation
End-to-end clinical system for speech transcription, medical translation, entity extraction, and FHIR-structured record generation. Benchmarked Qwen3 (1.7B/4B/8B) and Mistral-7B — Qwen3-8B won with 94% FHIR and 92% translation accuracy at 0.82s, with a self-hosted ASR reaching 91% clinical accuracy.
PROPITIVEAI Agent for Real Estate Agencies
A multi-agent LLM system with three autonomous agents for conversations, lead qualification, and scheduled outreach. Resolves tenant and landlord queries, escalating critical cases, with a hybrid lead-scoring system that cut cost by 50%.
QUERIVEMulti-Tenant AI Chat Backend
Production FastAPI backend powering an embeddable AI chat widget that answers customer questions and converts leads into tickets. Multi-tenant RAG with PostgreSQL, pgvector, BM25, and RRF fusion, achieving 94% accuracy.
Bone Fracture DetectionReport Synthesis & Localization
AI system that detects fractures from X-rays and writes the report — now in use at a hospital in Pakistan. Started with YOLO, moved to multimodal models, and fine-tuned a vision-language model to 82% accuracy on a real dataset, with lesion localization highlighting the exact fracture area.
Chughtai BotUrdu Voice Agent
Inbound calling agent that answers lab queries — reports, home sampling, doctor info — for Chughtai Lab. Built a Hindi-to-Roman pipeline to support Urdu, reaching 80% medically-aware accuracy, with tuned voice cloning for natural speech.
Gross Examination PipelineHospital Audio to Structured JSON
FastAPI backend that turns up to 4 hours of noisy hospital audio into structured JSON, built independently. Handles transcription, medical-term correction, and parsing to automate manual reporting.
MY JOURNEY
HowIGot Here
From a university thesis on anomaly detection to shipping AI in production hospitals — here is every chapter that got me here.
BS Software Engineering, COMSATS
Studied Software Engineering at COMSATS University Islamabad (CGPA 3.41/4.0). Thesis: self-supervised video anomaly detection using a memory-augmented LSTM autoencoder — reaching 0.9861 AUC-ROC on UCF-Crime, published on Zenodo.
Software Engineering Intern, COMSATS Software House
Completed a 3-month industrial internship focused on software development. Built CUI Timetable, a scheduling app that replaced paper-based timetables for faculty and students.
Associate Software Engineer (AI/ML), MDSOL Technologies
Worked as a full-stack AI engineer across frontend, backend, and AI integration. Built and shipped PROPITIVE and QUERIVE from prototype to production, designed multi-agent LLM systems and agentic RAG pipelines, and was recognized as Employee of the Month for AI contributions.
Software Engineer (AI/ML), BITLogix
Shipping AI features across four production healthcare products, including a fracture detection system and an Urdu voice agent. Deployed a self-hosted speech-to-text engine on vLLM, fine-tuned vision-language models, and led internal training on Docker and vLLM.
MY STACK
Poweredby Industry-GradeTools
I work across the full AI infrastructure landscape — picking the right tool for each problem, not the fashionable one.
CompaniesI've WorkedWith
From healthcare platforms to e-commerce brands — real projects, real results.
WhatOur ClientsSay
Don't just take our word for it — here's what people say after working with me. Real clients, real projects, real results.
Orhan N.
CEO, Adabassar Studios
DewByte built an automation pipeline that removed a ton of manual work from our ops. The workflow is stable, well-documented, and easy to extend. Communication was fast and proactive.
CERTIFICATIONS
Continuous Learning
AI moves fast — I keep pace with structured courses and specializations, not just tutorials.
PureLogics
Python + AI Bootcamp (Batch-5)
Python, ML, deep learning, RAG, and vector databases.
Let'sbuild somethingthat actuallyships together.
Whether it's an AI system, a voice agent, or a full-stack product — I'd love to hear what you're working on and see where I can help.