Deven
Liscombe
Profile
I build production AI systems end-to-end. Agent runtimes, memory layers, and product-facing infrastructure. Shipped a full platform as a solo engineer.
Open to remote-first roles at teams shipping AI products.
Readout
Focus
AI Systems
Infra + Product
Education
UWaterloo
BMath '25
Primary
TS / PY
Full Stack
Scope
End-to-End
Solo Builds
Experience
Founder / AI Systems Engineer
- 01Shipped working beta to 100+ MAUs with authenticated sessions and persistent multi-session agent state
- 02Designed DAG-based context graph architecture for long-lived agent memory, mitigating LLM token window constraints via structured graph traversal
- 03Built internal agent runtime supporting tool execution, memory writes, execution tracing, and MCP-compatible server for external agent integration
- 04Shipped consumer web app, cross-platform Electron desktop app, and Operator Studio SDK as sole engineer
Minto Group
2023Financial Analyst
- →Oversaw finances and cash flow projections for commercial real estate portfolio
- →Built VBA-automated budget templates, replacing manual workflows
Enwave
2022Corp Dev Finance Intern
- →Financial modelling for optimal pricing on historical electricity rate data
- →Rebuilt data archive into dashboard format for deal pricing
Education
University of Waterloo
BMath (Honours Mathematics)
2021 – 2025Major in Statistics.
Tech Stack
Languages & Frameworks
AI Systems
Also Used
Featured Project
Concurrent
Concurrent is a context operating system for AI. It unifies persistent memory, tool execution, and tracing into a single layer that any AI tool can plug into via MCP. Built on a knowledge graph that maintains structured context across tools, sessions, and tasks. One shared context layer instead of every tool starting from zero.
System Architecture
What Shipped
- 01Consumer web app (primary)
- 02Marketing website
- 03Desktop + mobile apps
- 04Operator Studio (SDK + UI)
- 01Context Graph memory system
- 02Agent Orchestrator (ReAct)
- 03MCP Server integration
- 04Backend APIs (REST)
- 01PostgreSQL schemas
- 02Vector embeddings (semantic)
- 03Persistent session storage
- 04Memory-scoped inference
Other Projects
Quantitative Trading System
Python · PostgreSQL · PyTorch · Monte Carlo · HMM · Event Streaming
Full-stack quantitative research and trading infrastructure. Hybrid regime classification, probabilistic risk modeling, and automated feature pipelines. Live trading consumes the same feature definitions as research, just computed in real-time. Same systems thinking applied to a different domain: data pipelines, state machines, and evaluation harnesses.
Data Layer
- 01Automated data ingestion and feature engineering pipelines
- 02Timeseries store for bars/ticks
- 03Corporate actions + calendar
- 04Versioned dataset snapshots
Research + Strategy
- 01Event-driven backtest engine
- 02Hybrid regime classifier (HMM + neural networks)
- 03Point-in-time correct ML training pipelines
- 04Evaluation harness across market regimes
Execution + Risk
- 01OMS with lifecycle state machine
- 02Monte Carlo simulation + Extreme Value Theory
- 03VaR and CVaR tail-risk estimation
- 04Broker reconciliation service
Contact
Open to remote-first roles on teams shipping AI products. Especially interested in agent infrastructure, developer tools, and AI-native workflows. Always happy to talk.