Available for AI Engineer & Applied ML roles · Munich / Hybrid

Mohammad Ayaz Alam

AI Engineer & Applied ML Developer

I build AI-powered software from model orchestration and evaluation to APIs, user interfaces, and cloud deployment. My experience spans industrial multi-agent systems at Siemens Energy, procurement automation at Zalion, and large-scale ML research across 90M+ records at the Institute for Employment Research.

Download CV (PDF) Portfolio PDF (4 Pages)
Mohammad Ayaz Alam
Location Munich, Germany
Focus Multi-Agent Systems & LLMs
Education M.Sc. Data Science (FAU)
Industry & Research Experience:
Zalion GmbH · Siemens Energy · BASF · Institute for Employment Research (IAB)
90M+
Records Analyzed
Engineered research ML pipelines at the Institute for Employment Research with positive-unlabeled learning.
2+ Yrs
Enterprise AI Systems
Production multi-agent frameworks, LangGraph orchestration, and human-in-the-loop validation checkpoints.
4+
Shipped Products
Founded Ilham Labs.ai: built Lattice Lab (multi-LLM canvas), Air Lune (real-time platform), and Ilmora.
M.Sc.
Data Science at FAU
Friedrich-Alexander-Universität Erlangen-Nürnberg. Focus on machine learning, time-series, and deep learning.
Featured Engineering Work

Selected Systems & Products

A curated look at multi-agent architectures, collaborative platforms, and on-device machine learning.

Featured Architecture Multi-Model AI Workspace
Lattice Lab

Lattice Lab: Parallel Reasoning & Canvas Workspace

A visual workspace bringing up to four language models onto a collaborative node canvas. Enables parallel prompting, conversation branching, model-to-model comparison, and stateful context merging.

  • Parallel model requests with real-time response streaming from 4 concurrent LLM providers
  • ReactFlow node-graph architecture with branchable conversation trees and canvas serialization
  • FastAPI backend with PostgreSQL/JSONB session persistence and authenticated workspaces
Next.js FastAPI ReactFlow PostgreSQL / JSONB Docker
Live App (lattice-lab.com) GitHub Repository
lattice-canvas // 4-model session
User Prompt
"Compare RAG vs Long-Context Fine-Tuning for high-compliance legal docs"
Claude 3.5 Sonnet
Focuses on deterministic citation, chunking provenance, and auditability...
GPT-4o
Analyzes latency trade-offs, KV-cache pricing, and multi-turn stability...
Live Production App Distributed Systems & Real-Time Audio
Air Lune

Air Lune: Shared Flight Multiplayer Focus Space

A multiplayer focus space that turns deep work into a shared flight between real airports. Participants join a shared cabin, follow live deterministic aircraft movement across a MapLibre map, talk over low-latency LiveKit voice, and stay accountable with server-authoritative focus timers.

  • Deterministic Geospatial Movement: Renders flight progress from persisted schedules using PostGIS instead of streaming raw coordinates over WebSockets, minimizing latency & bandwidth.
  • Browser-to-LiveKit WebRTC Voice: Direct audio channels without API proxying, plus local audio engines for ambient cabin, rain, and white-noise soundscapes.
  • Server-Authoritative Focus Engine: Redis & FastAPI backend with local countdown reconciliation, persistent flight logbooks, and private task isolation.
  • Production Deployment: Containerized with Docker Compose and Caddy on AWS Lightsail, with retained Terraform ECS/RDS scaling paths.
Next.js FastAPI PostgreSQL / PostGIS Redis LiveKit (WebRTC) MapLibre GL Docker AWS Lightsail Caddy
Open Air Lune (airlune.space)
MUC ✈ SFO
CRUISE · FL380
✈
Munich (MUC) 68% of route San Francisco (SFO)
Focus Timer 24:18 remaining
Voice Cabin ● LiveKit Active (4/4)
Cabin Crew Ayaz (Multi-Agent API), Sarah, Leo
Soundscape Cabin Hum + Rain
On-Device Edge AI Mobile & Privacy
Ilmora

Ilmora: Offline Android LLM Application

An offline mobile environment running quantized open-weights models (GGUF) directly on smartphone hardware. Zero data leaves the device, providing complete confidentiality and offline utility.

  • Hardware-accelerated llama.cpp inference engine compiled with Android NDK & CMake
  • Quantization-aware runtime handling with token-budgeted prompting
  • Encrypted local conversation storage with zero cloud telemetry
Flutter llama.cpp GGUF Android NDK CMake
Explore Ilmora
Agentic Workflows LangGraph Automation

Multi-Agent Project Generator

An autonomous development team powered by LangGraph. Deconstructs natural-language specifications into architecture blueprints, generates multi-file codebases, and validates syntax through typed agent handoffs.

  • Planner-architect-coder workflow with Pydantic typed schema validation
  • Stateful feedback loops for automatic compilation bug fixing and test scaffolding
  • Sandboxed file operations exporting production-ready project repositories
LangGraph Gemini / GPT-4 Pydantic Python
GitHub Repository
Zero-Knowledge Encryption Time-Locked Web App
When You Open Me

When You Open Me: Write to the Future

Write a message to the future, locked with a password and a date. No one - not even the server - can open it before its time. Encrypts messages and file attachments (up to 50 MB each) with scheduled delivery; recipients cannot read content until the release time arrives and they enter the correct password.

  • AES-256 Fernet Encryption: PBKDF2-HMAC-SHA256 with 100,000 iterations and unique random salt per message; passwords and plaintext are never stored.
  • Server-Enforced Time Locks: Database-level release verification with live countdown timers, email invitations, and unlock receipts.
  • Encrypted File Attachments: Attach up to 5 files (50 MB each) encrypted with the shared password before upload to S3, with an in-browser file viewer.
  • Full Self-Hosted Stack: FastAPI, PostgreSQL, Docker Compose, Nginx reverse proxy, scheduler, and Certbot automated SSL on AWS EC2.
FastAPI PostgreSQL AES-256 (Fernet) PBKDF2 AWS S3 Docker Compose Nginx
Open whenyouopen.me
3,600+ Downloads Model Fine-Tuning & RL

GPT-2 Poetry Generator with Reinforcement Learning

Fine-tuned a GPT-2 language model on poetry using custom reward functions for rhyme, meter, and stylistic coherence. Evaluated with policy gradient optimization (PPO) and open-sourced on Hugging Face.

  • Custom reward scoring evaluating phonetic rhyme density and syllabic meter
  • Reinforcement learning with policy gradients to steer generation quality
  • Published model weights with over 3,600+ downloads on Hugging Face
PyTorch Hugging Face GPT-2 Policy Gradients
Hugging Face Model (3.6k+ DL)
Applied ML & Data Automated Validation

Data Quality Assessment Application

An intelligent data profiling and auditing application. Instantly scans tabular datasets (CSV, Excel, JSON), evaluates completeness, consistency, and validity, and leverages LLM agents for executive diagnostic reports.

  • Automated statistical profiling with anomaly and outlier detection across schemas
  • AI-powered diagnostic recommendations for data pipeline remediation
  • Downloadable executive PDF audit reports for business and engineering stakeholders
Python Streamlit Gemini API Pandas Plotly
Try Live App
Data Quality Assessment Tool UI

Complete Project Archive

Browse all 25 production apps, AI experiments, research tools, and open-source packages.

Career History

Professional Experience

Track record across industrial AI engineering, research institutions, and product development.

Junior AI Engineer

Zalion GmbH · Munich, Germany
Mar 2026 - July 2026
  • Developing multi-agent procurement workflows using LangGraph and LangChain, integrating specialized agent endpoints into business chat workflows.
  • Engineering backend microservices with FastAPI and responsive user interfaces with Next.js.
  • Optimizing agent prompt reliability, output validation, and orchestration execution latency to ensure consistent business responses.
  • Contributing to containerized deployment with Docker on AWS EC2 and automated CI/CD delivery pipelines.
Multi-Agent Systems LangGraph FastAPI Next.js AWS EC2 CI/CD Docker

Master's Thesis · Junior Data Scientist, NLP

Siemens Energy · Erlangen, Germany
Jul 2025 - Feb 2026
  • Architected a modular multi-agent framework powered by LangGraph with decoupled orchestration, agent execution, tools, and persistent memory.
  • Built FastAPI streaming endpoints with OpenAI-compatible interfaces and PostgreSQL state checkpointing for auditable execution.
  • Engineered human-in-the-loop verification checkpoints for high-risk external actions (e.g. automated email dispatch, parameter modifications).
  • Applied the framework to two industrial production applications: Factory Assistant (engineering requirements parsing) and Terminal Management System Assistant (telemetry log analysis & automated diagnostics).
LangGraph PostgreSQL Human-in-the-Loop FastAPI Industrial NLP Log Diagnostics

Founder & AI Product Engineer

Ilham Labs.ai · Independent Initiative, Germany
2025 - Present
  • Leading product architecture, full-stack engineering, security, and cloud deployment across independent software products.
  • Shipped Lattice Lab (node-based multi-LLM workspace with parallel streaming and branching context) and Air Lune (multiplayer real-time focus platform).
  • Built whenyouopen.me (cryptographic time-locked messaging service with AWS S3 storage).
  • Managing containerized infrastructure with Docker, Nginx, AWS EC2, and S3.
System Architecture Next.js FastAPI Docker AWS llama.cpp

Machine Learning Researcher / Research Assistant

Institute for Employment Research (IAB) · Nuremberg, Germany
Nov 2024 - Feb 2026
  • Built scalable preprocessing and feature engineering pipelines handling over 90 million records of longitudinal labor market data.
  • Trained and benchmarked predictive ranking models using XGBoost, Random Forest, SVM, and positive-unlabeled learning (including Elkan & Noto).
  • Addressed extreme class imbalance and data sparsity to support empirical findings and decision support for federal research publications.
90M+ Records XGBoost Positive-Unlabeled Learning Feature Engineering Python

Data Consultant Intern

BASF · Ludwigshafen am Rhein, Germany
Apr 2025 - Jul 2025
  • Developed internal generative AI tools, including an automated Data Validation AI Agent providing structured checks, configurable tests, and quality scores.
  • Engineered the Inhouse Consulting event application with secure user authentication, submission tracking, live voting, and analytics.
Data Validation Applied Generative AI Internal Tools Streamlit
Core Competencies

Technical Expertise

Practical tools and frameworks I use to engineer robust AI systems.

AI & Multi-Agent Systems

  • LangGraph & LangChain (Orchestration, State Graphs)
  • Multi-Agent Systems (Human-in-the-loop, Guardrails)
  • Local LLM Inference (llama.cpp, GGUF, Quantization)
  • RAG & Vector Search (Chroma, Vector DBs, Hybrid Retrieval)
  • Model Context Protocol (MCP) & Function Calling
  • Fine-Tuning & Evaluation (RLHF, Policy Gradients, PPO, LoRA)

Machine Learning & Data

  • Languages: Python, SQL
  • ML Frameworks: PyTorch, Scikit-Learn, Transformers
  • Algorithms: XGBoost, Random Forest, SVM, PU Learning
  • Data Processing: Pandas, NumPy, Feature Engineering
  • Scalability: Large-scale Pipelines (90M+ records)
  • Analytics & BI: Power BI, Plotly, Seaborn

Full-Stack & Cloud

  • Backend: FastAPI, Node.js, REST APIs
  • Frontend & UI: Next.js, ReactFlow, TypeScript, Tailwind
  • Mobile: Flutter & Dart (Cross-platform)
  • Databases: PostgreSQL / JSONB, SQLite, Redis, Firebase
  • Cloud & DevOps: AWS (EC2, S3), Docker, Nginx, CI/CD
  • Version Control: Git, GitHub Actions, Linux
Spoken Languages: English (Professional Working Proficiency) German (A2 / Intermediate) Hindi & Urdu (Native Proficiency)
Academic Background

Education & Certifications

Rigorous foundations in computer science, statistics, and machine learning.

Degree Programs

Oct 2022 - Present

Master of Science in Data Science

Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Germany

Focus areas: Machine learning for time series, statistical learning theory, linked data, deep learning, business analytics.

Aug 2018 - Sep 2021

Bachelor of Computer Applications

Guru Gobind Singh Indraprastha University (GGSIPU), Delhi

Focus areas: Data structures & algorithms, relational databases, software engineering, computer networks.

Verified Certifications

AI Agents Fundamentals Hugging Face
Credential
ChatGPT Prompt Engineering for Developers DeepLearning.AI
Verify ↗
Full Stack Data Analytics iNeuron
Verify ↗
Deep Learning Foundation iNeuron
Verify ↗
Intro to Machine Learning Kaggle
Verify ↗
NLP Foundation iNeuron
Verify ↗
Let's Connect

Looking for an AI Engineer who ships?

I am open to full-time AI Engineering and Applied ML roles in Munich, across Germany (hybrid/onsite), and remote worldwide. Whether you are scaling an agentic workflow, optimizing local inference, or building high-volume pipelines, let's talk.

Direct Email
Documents & Profiles
Download CV Portfolio PDF LinkedIn ↗ GitHub ↗

Send a Quick Message