Dementia EEG ML Thesis
MSc Thesis · Imperial College LondonDeveloping closed-loop non-invasive brain stimulation intervention for dementia.
EEG time-series analysisLinear SVMAlzheimer's classificationBiomarker discoveryUp to 77.1% balanced accuracy
available for AI workLondon, UK
Viswash Mehta
AI EngineerGenAI ConsultantSystems Thinker
I design and build intelligent systems that turn complex enterprise knowledge into usable, scalable, and trustworthy AI products.
Currently building enterprise GenAI architectures across agentic workflows, multimodal retrieval, and knowledge discovery.
5+yrs building AI
10+systems shipped
Imperialapplied ML
system_boot.loglive
RAG
Agents
LangGraph
Vector DBs
Multimodal AI
Cloud AI
System Design
Product Thinking
02SKILL_CONSTELLATION
Skill Constellation
A connected map of capability — not a checklist. Select a cluster to zoom into the skills and the systems they power.
cluster // ai-systems
AI Systems
Designing the intelligence layer — retrieval, reasoning, and agentic orchestration that turns models into products.
GenAIRAGAgentic AILangChainLangGraphMCPA2A ProtocolPrompt EngineeringLangSmithVector Databases
powers // related systems
03MISSION_LOGS
Mission Logs
A record of the systems built and teams led — from research labs to enterprise GenAI delivery.
Leading GenAI solution design and implementation for enterprise clients, including scalable multimodal agentic Graph RAG architectures for operational intelligence and knowledge discovery.
Enterprise GenAI architecture
MultiModal Agentic Graph RAG
Team leadership
AI delivery at scale
Client-facing consulting
measurable_impact
- Architected agentic Graph RAG for operational intelligence
- Led GenAI delivery across enterprise client engagements
- Translated ambiguous business problems into deployable architectures
LangGraphGraph RAGMultimodal AICloud AISystem Design
04AI_ARTEFACTS
AI Artefacts
Selected systems, each engineered for real-world complexity. Open any artefact to explore its architecture.
05OPERATING_SYSTEM
How I Think
A personal operating system for building AI that survives contact with reality.
01process_01
Understand the Problem Space
Before building AI, understand the humans, workflows, constraints, and failure modes.
02process_02
Design the Intelligence Layer
Choose the right combination of retrieval, reasoning, agents, models, and evaluation.
03process_03
Engineer for Reality
Build systems that survive messy data, enterprise constraints, latency, cost, and trust requirements.
04process_04
Deliver with Ownership
Move from prototype to production with clarity, documentation, stakeholder alignment, and measurable impact.
05process_05
Keep Learning
AI changes quickly. The best engineers keep rebuilding their mental models.
06PORTFOLIO_AGENT
Ask My Portfolio
An AI interface trained on Viswash's profile. Ask a question or pick a prompt to begin.
Portfolio Agent
knowledge_base_onlineAsk about Viswash's systems, GenAI experience, leadership, or which projects fit enterprise AI.
Simulated agent · architecture is ready to connect to a real RAG backend over CV, portfolio, GitHub, and blog posts.
07ACADEMIC_SIGNAL
Academic Signal
Foundations in applied machine learning and computer engineering.
Oct 2022 – Oct 2024
Imperial College London
PGDip / MSc-level Applied Machine Learning
Applied ML, deep learning, and research-grade modelling.
London, UK
Nov 2018 – May 2022
Gujarat Technological University
Bachelor of Computer Engineering
Foundations in computer engineering and software systems.
Ahmedabad, India
08RESEARCH_FOOTPRINT
Research Footprint
Peer-reviewed and thesis-level work spanning neuroscience and computer vision.
A Survey on Vision-based Elder Fall Detection
Springer LNEEUsing Deep Learning Models.
Deep learningComputer visionElder safetyFall detection
09OPEN_CHANNEL
Start a Conversation
Let's build intelligent systems that actually work.
Whether it's GenAI consulting, intelligent product design, agentic systems, or enterprise AI architecture — I'm always open to meaningful conversations.
based // London, UKImperial College Londonsignal_detected