Contract software engineering
Portfolio and research site · papers, ML, and side projects
About
I am a software engineer working on full stack applications and AI/ML in production settings. Most engagements are about implementing features, tightening services that have outgrown their first design, or untangling data pipelines when reality stops matching the original diagram.
Energy, EV charging, and credit programs are where my recent work has landed. The technical patterns are similar across them: data arrives from different vendors and older schemas, identifiers are partial, and the business still needs one coherent view it can act on. Defining the normalization layer, the incremental loads, and the operator tooling around exceptions is the part I genuinely enjoy.
On regulated topics I work alongside the policy people and counsel who own the decisions, then encode those decisions in software that holds up under audit. Communication is a strength: short written updates, clear questions when something is ambiguous, and code that reads like the system it is supposed to be.
Engagements are usually remote and shaped to fit: a defined slice over four to twelve weeks, a focused tightening pass on an existing service, or a part-time retainer when steady commits matter more than a single push. You work with me directly.
Get in touch
Pick a time that works, or send a note first if that is easier.
Services
Each engagement ends in working software your team can run, with a written scope and clear acceptance checks agreed up front.
Walk through the codebase, map services and data flows, and surface the gaps that matter (tests, observability, schema drift). The output is a short written plan with sliced backlog, risks, and what I need from your domain owners.
User-facing flows, backing APIs, third-party integrations, migrations with rollback paths, and integration tests on the paths that touch money or regulated fields.
Ingestion for messy feeds, with validation rules, dead-letter queues, reconciliation reports, and operator screens that turn exceptions into something a person can act on.
LLM and classical ML features wired behind the auth model: prompts, tool contracts, evaluation hooks, rate limits, and human-in-the-loop review when outputs carry weight.
Expertise
Based in EST Timezone, working remotely with teams in compatible time zones. Strengths cluster around four areas, with the energy and credits work being the most recent context.
Features that touch PostgreSQL (and friends), batch and incremental jobs, API contracts, and deployment hygiene. At CFR Platform the work skewed database and pipeline heavy; at NuNet it leaned toward CRM-style TypeScript front ends talking to live services.
Model-backed behavior inside a real product: inference paths, prompt templates in code, structured outputs, evaluation datasets, and kill switches when confidence drops. Grounded in prior research on ML and identity-oriented AI.
EV charging and DER-shaped workloads treated as engineering problems: telemetry and billing feeds from many sources normalized into one service layer your product can rely on.
Workflows and data stores that match what your policy owners and counsel decide. When rules shift, code and tests move with them.
Clients
From recent projects with CFR Platform and NuNet through Riipen Level UP cohorts. Project specifics stay under NDA; names and roles are quoted as written.
Dylan combines deep technical skill with plain communication. Our brief was loose and our team needed direction. He tightened scope quickly, kept collaboration on track across contributors, and delivered documentation we could run with after handoff. We would work with him again.
Luc Lendrum Director of Operations, CFR Platform
Dylan showed up motivated and productive end to end. Clear ownership, quick cycles, and solid judgment on automation and workflow changes.
Luca Mazzo President, NuNet
Postgres masterclass, YAML files in perfect sync, Dylan steals the show.
Luc Lendrum Director of Operations, CFR PlatformSeparate endorsement from the same author, kept verbatim because the haiku was intentional.
Contact
A few sentences is enough to get a useful reply: the stack you run, where the data lives, what is slow or broken today, and whether you want a scope conversation first or hands-on time straight away. Replies usually go out within a few business days.
Portfolio and research site (papers, projects, code-adjacent work)
Email: dylan.todd04@gmail.com
Calendar
Pick a time that suits you. If nothing on the calendar lines up, send a note with a couple of windows that do.
Add your Calendly or Cal.com link in site-config.js (bookingEmbedUrl and bookingProvider).