Integrations & Internal Tools for Support and Ops Teams
I connect Intercom, Slack, Salesforce, Stripe, and the rest of your stack, and build the dashboards, secure logins, and in-house AI tools that take the busywork off your team. Senior engineer, custom code, fixed price per project.
Taking new projects.
Remote, UTC+5. Fixed price per project.
Custom code in Ruby on Rails, Go, Python, React, or Hotwire. No chains of
no-code automations that break quietly when one step changes.
Support tooling for Intercom teams
Conversations tagged and routed on arrival, account and billing context pulled into the inbox so agents stop switching tabs, escalations posted to the right Slack channel, and SLA risks flagged before they turn into churn.
In-house AI models
AI that runs on your own servers, so customer data never goes to a third-party AI provider. That matters for security, for your customers' privacy, and for rules like GDPR, HIPAA, and the data residency terms in your contracts. I fine-tune lightweight open-source models into experts on your subject, connect them to your internal tools, and debug them when the answers go wrong.
Integrations and two-way syncs
Intercom, Salesforce, Stripe, Slack, your own app, or any REST API. Verified webhooks, retries, and idempotent handlers, so a record is never written twice and a failed call never disappears silently.
Authentication and login security
Single sign-on over SAML and OpenID Connect with Entra ID, Cloudflare Access, or any identity provider. OAuth sign-in, passkeys, two-factor authentication, and bot protection with reCAPTCHA and Cloudflare Turnstile that stops fake signups without punishing real users.
APIs, clients and servers
REST APIs for your product with API keys, scopes, rate limits, and OpenAPI docs. OAuth provider or client, depending on which side of the connection you are. Clients for third-party APIs that handle pagination, retries, and token refresh.
Internal tools that replace spreadsheets
Admin panels, ops consoles, and approval flows with roles and an audit log. The spreadsheet three people edit at once becomes a tool with one source of truth.
Dashboards and reporting
The numbers someone assembles by hand every Monday, live on one screen, plus scheduled digests to Slack or email.
Security built in, not bolted on
My background is security engineering. Least-privilege API scopes, secrets kept out of code, signed webhooks, and audit trails come standard, because your tools touch customer data.
Recent work
Client and employer names are left out. Happy to walk through any of these on a call.
Subject-expert AI on modest hardware · In-house AI, self-hosted
Lightweight open-source models - including IBM's Granite 4.2 and other open-weight models in that size class - fine-tuned into experts in one specific field, so they run on-prem on modest hardware instead of an expensive general model. Tuned the output language so answers read naturally to customers and improve their experience.
Company-wide authentication platform · Top-10 global certificate authority
Sole architect of login and identity for tens of thousands of daily users: OAuth2, OpenID Connect, SAML, external identity providers, and the OAuth provider side, not only the client. Extracted authentication into its own service with sessions that work across services, and closed account-takeover and second-factor-bypass paths.
Bot protection at 1,000+ fake signups a day · Top-10 global certificate authority
Layered signup defense with reCAPTCHA v2/v3 and Cloudflare Turnstile, plus countermeasures for human CAPTCHA farms that solve challenges by hand. Fake accounts stopped without adding friction for real customers.
Stripe billing that cannot double-apply · SaaS product, Ruby on Rails
Full subscription lifecycle - checkout, upgrades, cancellations, refunds, failed payments, trial reminders, and disputes - driven by eight Stripe webhook events. Every event is recorded before it is processed, so Stripe's retries never charge, refund, or downgrade anyone twice.
An MCP server and Claude plugin for a live product · SaaS product, Ruby on Rails + Python
Seventeen tools that let Claude work inside a user's own account, behind OAuth with PKCE, dynamic client registration, a mandatory consent screen, and the same read/write/admin scopes as the REST API. Shipped as a published Claude Code plugin with slash commands, skills, and an eval suite run before each release.
One business layer, three surfaces · Standalone service, Go
A Go service where the web UI, the MCP endpoint, and a JSON API all run through one tool registry, so they cannot drift apart. Callers subscribe to state changes through HMAC-signed webhook callbacks, and every state change is written to an event log in the same transaction.
Team alerts where people already work · SaaS product, Ruby on Rails
Per-team notifications over Slack, email, and custom webhooks. Destination URLs are resolved and checked before use, so a customer-supplied webhook cannot be pointed at internal infrastructure.
Admin console and email pipeline · SaaS product, Ruby on Rails + Hotwire
Internal console for users, teams, subscriptions, capacity, and monitoring stats. Moved the newsletter off a third-party email platform onto Amazon SES, with bounce and complaint handling over verified SNS notifications.
Staff tooling for fraud and abuse response · Top-10 global certificate authority
Internal detection and review tooling that pairs LLM-assisted pattern matching with human-led forensics, taking the team from first signal to mitigation in minutes instead of days. Also covered staff and admin activity, not only customer accounts.
How it works
1.
Tell me what your team does by hand
A short message, or a call if it helps. Which tools, which steps, and how often.
2.
Written scope and a fixed price
What it will do, what it will not do, and how we will both know it is done. You pay per project, not per hour.
3.
Built in your accounts, demoed as it grows
Code lives in your repository and credentials in your accounts from day one. You see working pieces at each milestone, not a big reveal at the end.
4.
Handoff you can run without me
Setup notes, a runbook for when something breaks, and a fix window after launch included in every quote.
Not every problem needs a full project. Sometimes your team needs a senior engineer
to look at the plan, point out the risks, and say what to do next. My goal for every
engagement: pay me $1, save $2.
In-house AI models
Running AI on your own infrastructure instead of sending data to an outside provider, for security, privacy, and compliance. Choosing the model, fine-tuning it, debugging it, and connecting it to your systems.
APIs and integrations
How your systems should talk to each other: API design, connecting tools like Stripe, Salesforce, Intercom, and Slack, and webhooks that never lose data.
Security and fraud
A review of your login, account recovery, and payment flows, and a plan to stop fake signups, account takeover, and abuse.
Authentication and identity
Choosing and rolling out SSO, OAuth, passkeys, and 2FA without locking out your real users.
PKI and certificates
Certificate lifecycle, automation, and what new browser and industry rules mean for your systems.
Performance and scale
Finding out why the app is slow, what to fix first, and what will break at the next 10x.
Ways to work together
One-off review
A written report with what I found, why it matters, and what to do next.
Monthly advisory
Regular calls with your team, plus questions by email or Slack between calls.
Hands-on help
I work with your engineers on the fix, not only on the report.
Contact
Tell me what your team does by hand.
Send a short note about the tools involved and the steps someone repeats every week. You get a written scope and a fixed price back.