Available for freelance & full-time AI roles

Muhammad Ahmad Mujtaba

AI Engineer·Deep Learning·NLP·MLOps

I build production AI systems end-to-end, from raw data to deployed, optimized services.

3 yrs
Production AI experience
Solo
End-to-end, international clients
3.61
B.Sc. AI · CGPA / 4.00
01 / about

Who I am

AI engineer with three years of experience building production AI systems: LLM fine-tuning, RAG pipelines, and full-stack ML backends with complex optimization layers. I've worked as the sole engineer on end-to-end projects for international clients, handling everything from raw data to deployed services. I hold a B.Sc. in Artificial Intelligence.

I care about the parts that don't fit on a slide: handling sparse and messy data, cutting inference latency, indexing high-dimensional vectors, and shipping services that hold up under real load, not just notebooks that run once.

B.Sc. Artificial Intelligence
COMSATS University Islamabad
Sep 2022 – Jul 2026 (Completed)
CGPA: 3.61 / 4.00
Remote · Pakistan · working with international clients
02 / stack

Tools I build with

The technologies behind the systems below, grouped by where they sit in the pipeline.

AI / ML
PyTorchHugging Face TransformersLangChainLightGBMTensorFlowYOLOStable DiffusionAgno
Techniques
LLM Fine-tuningLoRA / QLoRARAGNLPComputer VisionPrice OptimizationTime-Series (STL)
LLM APIs & Local Inference
OpenAIAnthropicGeminiAzure AIllama.cppOllama
Languages & Web
PythonJavaScriptFastAPI
Data & MLOps
DockerMongoDBQdrantSupabaseGitDVCGCPDigitalOceanLinuxn8nPrefectJava
03 / experience

Problems, and how I solved them

Not a list of job duties. Each item is a real constraint I hit, what I did about it, and the result it produced.

Feb 2026 – Present

Contract AI Engineer

HKS TechLabs
Remote
  1. constraint

    A client needed data-driven pricing across many SKUs, with explanations they could trust.

    approach

    Sole ML engineer building the pricing recommender from the ground up: Supabase ingestion → SKU classification → STL decomposition.

    result

    A full pipeline carrying raw catalog data through to clean, model-ready demand series.

  2. constraint

    Forecasting alone doesn't set a price; the recommendation has to be defensible per SKU.

    approach

    LightGBM demand forecasting (Tweedie + quantile regression), elasticity simulation, and a two-phase price optimizer (LP + SLSQP).

    result

    A production price recommender with per-SKU explainability behind every recommendation.

Jun 2025 – Jan 2026

Contract AI Engineer

SaaS Startup (client name withheld under NDA)
Remote
  1. constraint

    A RAG recommender was embedding raw objects, making retrieval slow: 15–20s on short queries and 50–60s on larger ones.

    approach

    Designed embeddings around semantic taxonomies instead of raw objects and re-architected the Qdrant retrieval layer.

    result

    Cut query times to 5–7s (short) and 10–13s (large), roughly a 3–5× speedup.

  2. constraint

    A monolithic setup capped throughput at 4–6 concurrent users.

    approach

    Migrated the same system to a microservice architecture with async request handling.

    result

    Scaled to 40–50 concurrent users.

  3. constraint

    Source data was messy, unstructured, and scattered across the web.

    approach

    Built end-to-end data pipelines: web scrapers, LLM-based JSON cleaning/structuring, and entity extraction (skills, education, tasks) before embedding.

    result

    Clean, structured, embeddable data feeding the retrieval system reliably.

May 2024 – Jun 2025

Junior AI Engineer

EnDevSols LLC
  1. constraint

    Training was manual and didn't reproduce reliably.

    approach

    Designed and maintained automated training pipelines on DigitalOcean and physical GPU servers.

    result

    Repeatable, hands-off training runs.

  2. constraint

    Clients needed custom NLP, generative, and computer-vision models behind real services, not off-the-shelf APIs.

    approach

    Built and deployed NLP, generative AI, and computer-vision models behind APIs, including object detection, classification, and LoRA/QLoRA fine-tuning workflows.

    result

    Shipped tailored models into client products.

  3. constraint

    Products needed retrieval-augmented and chained-LLM features.

    approach

    Contributed LangChain and RAG-based product integrations alongside the core engineering team.

    result

    RAG features integrated into shipping products.

Jun 2023 – Apr 2024

AI Engineer Intern

EnDevSols LLC
  1. constraint

    Early-career: learning to ship ML into real services.

    approach

    Trained and deployed models, handled data preprocessing, integrated vector databases, and built APIs with FastAPI.

    result

    Foundation in the full model-to-API lifecycle.

04 / impact

The numbers, and why they were hard

Real outcomes from production work. The pricing curve is illustrative; the latency and scaling figures are from shipped systems.

RAG retrieval latency

3–5× faster queries

Embedding around semantic taxonomies instead of raw objects, plus a re-architected Qdrant layer.

Before After
Concurrency

Monolith → microservices

Async request handling lifted a hard concurrency ceiling without rewriting the models.

0
before
0
after

concurrent users

Profit optimizer · illustrative

Finding the profit-maximizing price

The optimizer doesn't scan every price. A two-phase LP + SLSQP search explores a bounded band around each SKU's current price (shaded), under margin floors and competitor caps, to maximize profit, not revenue. Demand falls as price rises; profit peaks where volume and margin trade off, and turns negative below unit cost.

Demand Profit
optimum ≈ $40 → $1,400 profit
05 / work

Selected work

Projects I can show in full, and client work I can only describe at a high level.

Final Year Projectview
Case study

IHS

Intelligent Home Surveillance: real-time fall detection, hazard recognition, child-safety monitoring, and RAG-powered first-aid guidance.

YOLOv8-PoseTCNFastAPIRAG (FAISS + BM25)FlutterSupabaseFirebase FCM
Personal ProjectIn Progressview
Case study

Demand-Forecasting Agent

Agentic ML pipeline for retail demand forecasting on M5-format data. Deterministic branch logic, data-relative thresholds, and a critic that gates acceptance — not a rubber stamp.

LightGBMWMAPERolling-origin CVTweediePythonpytest
Client work · NDA

Taxonomy-Driven RAG Recommender

Semantic-taxonomy embeddings on Qdrant with async microservices: 3–5× faster retrieval, scaled from 4–6 to 40–50 concurrent users.

QdrantRAGMicroservicesAsyncLLM pipelines
Client work · HKS TechLabs

AI Profit Optimizer

Production pricing engine for an e-commerce catalog. LightGBM demand forecasting (Tweedie + quantile) over STL-decomposed history feeds a two-phase LP + SLSQP profit optimizer bounded by margin guardrails and competitor-aware caps. Adds champion/challenger model gating (WMAPE), a self-tuning feedback loop, and DB-backed runtime policy across sell- and buy-side pricing.

LightGBMSTLLP + SLSQPChampion/ChallengerSelf-tuningSupabase
Client work · EnDevSols LLC

Generative · NLP · Vision Models

LoRA/QLoRA LLM fine-tuning, custom CV/NLP models, and automated GPU training pipelines delivered into client products.

LoRA / QLoRAFine-tuningComputer VisionDigitalOcean
06 / contact

Let's build something

Open to freelance projects and full-time AI roles. The fastest way to reach me is email.