Siddarth Dagar

I build infrastructure for engineers: platform systems, developer tooling, AI/LLM workflows, type-driven product architecture, and performance work across large production codebases.

Recently that has meant structured backend/entity work, schema and runtime-type infrastructure, backend hot reloading, distributed agent orchestration, LLM alert workflows, customer-critical product migrations, and performance debugging.

Reach me by filling in the blank: sidmcheath_gmail.com.

Selected Work

First-class training/quiz infrastructure inside a structured entity platform, including schema, permissions, cached parsing, and builder/learner modes.
Generic type patterns in a production schema system, extending parser, runtime metadata walking, spec definitions, and builder validation.
Backend hot module reloading for web, ML, asynchronous job, and LLM server routes.
Agentic LLM alert workflows from model plumbing and tool execution to structured recommendations and chat handoff.
Distributed AI-agent orchestration over Kubernetes Jobs, NATS JetStream, PostgreSQL, and Redis.
Page-load, PostgreSQL, Git, and legacy frontend performance work in a mature product surface.

Other cool stuff I’ve done.

zpaq and Asteroids in MIPS assembly.
tetra.c, a spinning ASCII tetrahedron.
A small fluid simulator in C++.
A 3D ResNet reproduction on video data.
C. elegans simulation experiments with PyTorch/RL.
A D&D planner and a Godot roguelike.

Recently

Shipped 7 production projects and 155+ customer/incident fixes across platform infrastructure, foundational product features, customer-priority workflows, and developer-experience systems.
Built a hierarchical training module MVP in under two weeks: module → page → block → question → answer/response.
Added dynamic parametrization to a cross-language type system that powers generated Python types, TypeScript types, API clients, runtime metadata, and dynamic UI surfaces.
Designed and shipped LLM-enabled alerting workflows that analyze platform data, fire structured recommendations, and hand users into contextual chat.
Migrated customer-critical process workflows from tree models to DAGs with edge entities, workflow-aware traversal, and graph UI rendering.
Worked on distributed AI-agent infrastructure where external events trigger ephemeral Kubernetes workloads coordinated over NATS JetStream.

Engineering Notes

Types are infrastructure

In a large product, schemas are not just correctness tools. They become APIs, SDKs, runtime metadata, form renderers, validation layers, permissions, and migration boundaries.

Schema YAML source of truth

Generator parser, gates, metadata pass

Typed surface Python, TypeScript, API wrappers

Runtime tools forms, pickers, validation, docs

One declaration becomes compile-time types and runtime behavior.

Recovery paths matter more than happy paths

Event-driven systems look straightforward until pods die, consumers get stuck, messages are replayed, or state needs to be reconstructed after restart. Good infrastructure makes recovery boring.

External event Slack, GitHub, Jira, web

Durable queue input, ack, replay state

Control plane reconciles queue, pod, and status signals

Worker does work, emits lifecycle and outcome

The useful system is the loop, not the happy-path arrow.

Developer experience compounds

A repeated workflow does not need to look important to be worth optimizing. Backend HMR, Git performance, and fast local loops create leverage because every engineer pays that cost every day.

AI-assisted engineering needs boundaries

AI tooling is useful for parallelizing boilerplate migrations and frontend iteration. Human judgment should own schema design, security models, parser/cache correctness, and product constraints.

Experience

Uncountable · Software Engineer

October 2025 -- Present

Platform engineering work across structured backend systems, TypeSpec/runtime metadata, developer experience, LLM/agent infrastructure, performance optimization, and customer-critical product workflows.

MathWorks · Engineering Development Group Intern

May 2023 -- August 2023

Built MATLAB code-intelligence tooling, TypeScript indexing, dataset-labeling integrations, and compilation workflow optimizations for internal developer tools and early LLM-powered assistant infrastructure.

Toronto Early Cognition Lab · Software Developer

September 2022 -- December 2022

Independently delivered a full-stack video-processing system for secure upload, annotation, anonymization, permissions, audit logging, and researcher handoff.

16Bit AI · Software Engineering Intern

May 2022 -- August 2022

Stabilized production medical-imaging model workflows, fixed cross-browser EXIF preprocessing failures, and supported real-time patient submission, validation, and clinical audit infrastructure.

About

I studied Computer Science and Mathematics at the University of Toronto. I am drawn to large systems, developer-speed infrastructure, scientific/product platforms, and tools that turn hard-won engineering patterns into reusable leverage.