Jamie Costello

Director of Product

Ratehub.ca

Oct 2021 - Present

Head of product & member of the executive leadership team setting the product vision for Canadas largest financial comparison platform.

Leadership & Mentorship

• Led Product team of 8 (PM & Design) driving results like growing insurance revenue 5x in 3 years by championing user-obsessed culture of utilizing feedback, experimentation, and analytics tools (Heap, GA, Looker) to drive product decisions and measure success
• Established an APM program at Ratehub to give opportunities to people struggling to enter Product, transitioned all 3 APMs into full-time PM roles, with 1 becoming a Senior PM within 2 years, by providing 1:1 coaching and mentorship

Data Driven Product Development

• Reduced friction for credit card applications showing a 3x lift in completion by leading the discovery, pitching & onboarding with our partners for our hosted Credit Card Application flow
• Executed 0-1 launches for multiple new verticals without increasing costs or scaling teams, driving net new revenue by launching lean POCs and taking data-driven decisions on where to further invest based on results
• Led the company-wide SEO strategy pivot, reclaiming the #1 SEO position for key high-intent landing pages by aligning product, development, content, and marketing teams around optimizing for user engagement metrics

AI & LLM Integration:

• Increased sales team productivity, freeing up 5-10 hours per week previously spent on manual data entry by collaborating with the head of sales to build LLM based tool to automate partner updates.
• Eliminated reliance on an expensive external vendor & reduced data error rate to nearly 0% by building a AI powered rate scraper
• Established a comprehensive internal documentation culture with 90% of key processes documented within 6 months by implementing an AI-powered system to automatically generate documentation from recorded meetings

Senior Product Manager, Everyday Banking, MoneySense & Frontend Infrastructure

Ratehub.ca

Aug 2018 - Oct 2021

Joined as PM and moved to a Senior PM position in 2020. Managed various scrum teams (Engineering, QA, design, analytics), responsible for defining the outcomes and metrics that would help us drive revenue & the roadmaps to get us there, primarily focusing on our credit cards business.

• Grew credit card revenue 4x by utilizing feedback from user interviews and surveys, analytics insights & experimentation to optimize all high intent landing pages, paid ad campaigns and key journeys
• Enhanced Credit Card application approval rates, achieving a 3x lift in revenue per conversion by focusing on approval predictors with Eligibility Checker tool. Based on MVP feedback, pivoted the project's initial focus of personalization to enhancing user approval confidence, substantially increasing engagement and revenue impact
• Significantly grew the reach of the Credit Card business, almost doubling the target audience and revenue of the vertical by seamlessly integrating Ratehub's optimized credit card journeys into the acquired Moneysense platform
• Delivered a high-performing, user-centric complete rebrand as measured by a 10% decrease in bounce rate and 30-point increase in Lighthouse performance by leading a cross-functional effort to redesign and rebuild Ratehub.ca in React

Account Security & Loss Prevention Manager

Tucows

Jan 2016 - Aug 2018

Promoted to lead a fraud prevention team across all major Tucows brands (Ting, Enom, Hover & OpenSRS), I was responsible for developing and implementing comprehensive security strategies while staying hands-on with operational tasks.

• Grew net revenue with loss prevention, with results like reducing revenue loss by 50% on Ting brand by analyzing data to identify trends and using those insights to automate workflows and optimize cross functional processes
• Rapidly integrated successful security overhaul at newly aquired ENOM, cutting chargebacks by 95% within 6 months by collaborating with product and dev teams to train & deploy machine learning based security tool, Sift
• Enhanced security & prevented abuse and phishing scams, driving a reduction of chargebacks of 75% for the Hover brand by overhauling processes & implementing new automated security systems

Fraud Prevention & Orders Specialist

Tucows

Oct 2013 - Jan 2016

Starting as a Customer Service Representative and quickly moving to a fraud prevention role for Ting, I contributed to the brand's growth while significantly reducing fraud-related losses.

• Prevented significant abuse on any launched marketing campaign, including including recovering approximately $65,000 in misused credits during one promotional period by using data to analyze activations for suspicious patterns
• Wore many hats during period of extreme growth, including coordinating with our new warehouse to implement an RMA system as we launched our own phone shipments to creating an internal tool to quickly check the validity of serial numbers on potential new stock purchases
• Partnered with Support, Product Management, Marketing, and Development teams to create processes that minimized liability and enhanced customer experience, such as payment extension handling

External Mentoring

Toronto Product Management Association

• Taking part in the TPMA mentorship program, providing one on one coaching and mentorship via zoom and in person

Youth Employment Services

• Worked with fellow Ratehub employees to regularly teach a financial literacy course and provided some one-on-one career guidance for their members

FeedBrief

FeedBrief is a pre launch, alpha stage AI Product I am building just as a proof of concept and not intending to launch to a wide audience. On the backend, It uses various specialized, small Large Language Models (Llama 3.1, Gemini Flash, GPT-4o-mini & Claude Haiku) to automatically scrape news articles on websites, create a bullet point summaries of the content, categorize the content and apply a relevancy score for each category. It runs in real time on a Ubuntu server and stores this data in a Supabase DB.

The frontend is a custom built NextJS project built out using TailwindUI which takes the data from that Supabase DB and turns it into news feed. Right now, the frontend and UI are very basic, but the goal is to turn it into something where you can view articles by topic, have them sorted by the relevancy score on that topic or have a personalized home feed with only the topics you care about personalized by your preference.

All code has been put together by me leaning heavily on the latest Large Language Models. I am not a capable developer, but all this was made possible by leveraging tools like Cursor to produce the bespoke code for me.

AnyBlock To Markdown Converter

Anyblock To Markdown Converter is an open source tool used to convert Anyblock files into universal markdown format. Anyblock files are a proprietary file type used by the note taking app Anytype. The issue was it is their default exported format, Anyblock, was a very verbose and difficult to decipher where one note could have 20+ JSON files with thousands of lines due to how complex a tool Anytype is.

I used various Large Language Models to piece together the Anyblock format and understand how the relationships work between each file, and from there created a mapping for how to turn Anyblock into Markdown. I then used the latest Large Language Models to turn that into a Python based tool that can automatically turn any Anyblock export into Markdown file or files.

The project is Open Source and was shared with the Anytype community and development team, and has received good feedback from both. My next step is to turn it into a packaged application so that people who aren't comfortable with terminals can also take advantage of it.

Meeting Transcriber

Meeting transcriber is a simple, open source tool which can take any audio or video recording of a meeting or presentation and turn it into a structured output of your choosing. Typical use cases would be meeting summaries in a format you like (eg markdown for something like Notion, or outliner format for something like Logseq) or to turn instructional or onboarding videos into searchable, editable entries in a company Wiki.

It works by using Whisper by OpenAI to produce a full transcription of the video or audio file, and it then sends that transcription to a Large Language Model to convert that into whatever format you want. You can have preset system prompts to cover common uses cases which you can choose from - so in the example above, a preset for meeting transcribing and one for technical documentation where you can define the structure and format of the notes.

It's aimed at anyone, but it is something Product Managers could find especially useful. You can use it to build out the technical documentation that may be the backbone of a project, or even use it to dictate long form, unstructured thoughts and turn those into the beginning point of a PRD. As you can use completely local models for both the transcription and the summary, it is also possible to use this tool while maintaining complete control over your private data.

Next steps are adding support for Faster Whisper for less powerful machines, and the option to use a cloud based transcription option for those who don't want to tax their machine at all. After that, I'll turn it into a packaged application so that people who aren't comfortable with terminals can also take advantage of it.

Github

To see these projects in action, or look at any other projects I'm working on, please feel free to visit my Github