Ka Lun Chan (KC)
Engineering practice
I build products and the teams that scale them.
Ka Lun Chan (KC) is a San Francisco Bay Area engineering leader and generalist with 20+ years across every layer of a business, from helpdesk, network operations and security to product, marketing and architecture. As co-founder and CTO, he built a startup from the ground up to 400,000+ users and an acquisition, leading a team of 30+ engineers, VoIP specialists and designers. He now leads government technology work through Yippify.
Start a conversation Experience Leadership
- 20+ years in software
- 400,000+ users scaled as co-founder and CTO, through an acquisition
- 30+ engineers, VoIP specialists and designers led
Founder story: from an idea to an acquisition
From carrier networks and physical data centers to cloud infrastructure, distributed systems and building businesses, I have spent more than two decades making technology work at scale, balancing performance, reliability, capacity and cost. Still hands-on.
- 2,000+ collocations, carrier capacity planning
- 8M+ VoIP minutes a month, server farm and expansion
- +50% organic and social traffic as publishing CTO
- ~30% lower AWS costs on a re-platform
Every layer of a company, one chapter at a time
Recently shippedFedPathBerkeley OmniumUnion City Smog Check
Teams that make decisions, solve problems and keep improving without depending on me.
Owned the outcome, not only the system.
- Co-founded a communications platform: 400k+ users on three continents, revenue in under four months, acquired
- Publishing: traffic and engagement up 50%, AdSense, affiliate and SEO as a business
- Six startups helped from idea to working product; FedPath, VeloWise, SurgeIQ and Useful Little Tools shipped
Boundaries that follow the business and fail in ways you can reason about.
Still in the code: prototypes, reviews and the hard parts of the build.
Capacity planned from measured demand, at a cost the business can carry.
Deny by default, allow explicitly, and test the rule with the traffic it should permit.
Where capacity was a purchase order and reliability was a process.
Seven layers, business at the top, wire at the bottom. Select one for the work behind it.
What the work produced.
Six things a hiring manager can check. Each links to the page that backs it.
- 20+years in technology, from the helpdesk to the boardroom
- Acquiredco-founded a communications company and led it to acquisition, staying on the board
- 400,000+users served from cloud regions on three continents
- 30+engineers, VoIP specialists and designers led, including fully remote teams
- ~30%lower AWS costs on a publishing re-platform, with traffic and engagement up 50%
- Federalapplications engineered to security authorization, Section 508 and Plain Writing requirements
Seven stages, one direction.
Stages, not dates. Select one for the role, the technology, what it produced and the lesson carried to the next.
Member of technical staff, national broadband network
What it produced
Integration, expansion and capacity planning across 2,000+ collocations; operations for network security and IP services.
Technologies
Cisco routing and ACLs, ATM, IP and VoIP, SNMP and Cacti, Expect and shell, an R forecasting program
Lesson carried forward
Measure first, plan from data, and order ahead of the lead time.
Case study: capacity across 2,000+ collocationsOne of the first three in a new Security Operations Center
What it produced
Supported the launch of a managed firewall and VPN service, built its operational process, and trained the Technical Assistance Center on the East Coast.
Technologies
Cisco ACLs, NetScreen firewalls and IPsec VPNs, managed security services
Lesson carried forward
A rule that blocks real traffic is an outage with a security label on it.
Network security long before DevSecOpsVP of Operations, voice services, then co-founder
What it produced
Ran the server farm behind 8M+ VoIP minutes a month, planned expansion ahead of demand, and built least-cost and geolocation routing with in-house billing.
Technologies
Asterisk, OpenSIPS, MediaProxy, SIP, Python AGI, C++ on Nokia phones
Lesson carried forward
Quality, cost and security are one decision; the cheapest route is often cheap for a reason.
What VoIP taught me about distributed systemsCo-founder and CTO, global communications platform
What it produced
400,000+ users across South America, Asia and Africa, revenue in under four months, a team of 30+, multimillion-dollar growth and an acquisition with a board seat.
Technologies
A multi-tier web and mobile platform, a data center cabled by hand, then cloud regions on three continents
Lesson carried forward
Build for the person using the system, not for the engineers building it.
Founder story: from an idea to an acquisitionCTO, media publishing; architect on client platforms
What it produced
Re-platformed a publishing portfolio from WordPress to Rails on AWS, with organic and social traffic up 50%; event-driven services on Kafka; APIs with authorization enforced on the record.
Technologies
Ruby on Rails, Python and Django, React and Next.js, PostgreSQL, Redis, Kafka
Lesson carried forward
A boundary should pay for itself, and the rent is due monthly.
Architecture lessons from very different systemsCTO, VP of Operations, interim head of engineering
What it produced
Led engineering at a startup through acquisition, at a media publisher, and for client teams, including fully remote ones; mentored engineers into leads.
Technologies
Decisions in writing, small pull requests, frequent deploys, security and accessibility as release gates
Lesson carried forward
The measure of a lead is the team that works without them.
How I run engineeringConsultant and founder, Yippify
What it produced
About 30% lower AWS costs on a re-platform; federal applications built to security authorization and accessibility requirements; FedPath and Useful Little Tools shipped with AI assistance.
Technologies
AWS with ECS and Fargate, Lambda, RDS, Terraform, GitHub Actions; LLM document processing; an AI assistant in the terminal
Lesson carried forward
Capacity became a bill, and the tools keep changing. Keep learning.
Building software before and after AIProblem, analysis, decision, implementation, outcome.
Five pieces of work in the same five steps. Open one, then read the full account.
- ProblemLinks filled months before replacement hardware could arrive
- AnalysisInterface statistics over SNMP; the peaks, not the average, told the story
- DecisionA headroom threshold decided in advance, per link
- ImplementationAn R program forecasting when each link would cross it
- OutcomeUpgrades ordered ahead of lead times instead of after congestion
- ProblemThe cheapest carrier route often failed to connect or sounded bad
- AnalysisAnswer rates, call durations, completion and quality by route and destination
- DecisionBalance cost against quality; choose media paths by caller location
- ImplementationCall statistics fed into routing; source-IP geolocation for the media path
- OutcomeBetter call quality for distant users and less bandwidth paid for
- ProblemA portfolio of titles competing for audience against far larger publishers
- AnalysisTraffic sources, rankings, engagement and revenue per thousand views by page
- DecisionTreat speed, structure and distribution as platform features
- ImplementationRe-platform from WordPress to Rails on AWS; structured, answerable pages
- OutcomeOrganic traffic, social traffic and engagement each up 50%
- ProblemHosting sized for a peak that never came back, on a stack that was hard to change
- AnalysisUtilization, instance sizing, database load and the pages that never changed
- DecisionRight-size, scale on real demand, fix the database before buying a bigger one
- ImplementationCaching, query and connection fixes, unused resources removed, Terraform-managed infrastructure
- OutcomeAbout 30% lower AWS costs, with faster pages
- ProblemApplications that must be demonstrably secure, accessible and maintainable
- AnalysisWhich controls the platform inherits and which the application owns
- DecisionSecurity, Section 508 and Plain Writing in the acceptance criteria, not a launch-week audit
- ImplementationServer-side authorization, audit logs, infrastructure as code, a pipeline that leaves evidence
- OutcomeEngineering evidence that supports the Authority to Operate
How I run engineering, as a loop.
Business goals in, measured outcomes out, and the measurement feeds the next round. Each step links to the practice behind it.
- Business goals
Start from the outcome the business needs, in its own words, and name the unknowns.
- Technical strategy
Architecture decisions in writing, with the options, the choice and what it costs to run.
- Team alignment
Every engineer can say why their task matters; product, QA and leadership are in the same loop.
- Execution
Small stacked pull requests, frequent low-risk deploys, security and accessibility as release gates.
- Measurement
Lead time, change-failure rate, sprint carryover and the infrastructure bill, read every cycle.
- Continuous improvement
Pay down the debt that slows the roadmap, and grow engineers into leaders so the system improves without me.
What the measurement shows goes back into the goals. The loop is the job; no single step is.