Focal Forensics

Building a Custom AI Solution for a Non-Technical Company

 
Crafted Focal Forensics Redaction Software

Focal Forensics is the leading provider of redaction services to law enforcement agencies in the United States. During a critical incident, a large collection of media is typically produced through the use of body-worn cameras, security footage, audio interviews and still images.

The video, audio, and images related to a case must be meticulously redacted with 100% accuracy to ensure privacy of unrelated parties before releasing publicly. Crafted was able to deliver a proof of concept (POC) in one week, which turned into a six week (and ongoing) engagement to build a custom internal tool to help automate and identify subjects that are in need of redaction. 

In parallel, we evaluated the market for artificial intelligence (AI) video redaction tools to make sure it was necessary to build a custom solution. After determining we needed a custom software solution, we codified Focal Forensics’ standard operating procedures into a secure, cloud-based tool in six weeks while implementing AI assistance. Steve Bufalino, Co-Founder of Focal Forensics, said, “The internal tool is being operationalized and we project significant bottom line impact from the investment.”

Goals

Focal Forensics brought in Crafted to create a tool that leverages the most advanced facial detection machine learning (ML) algorithms to help detect and redact sensitive information. This advanced capability would reduce time spent on redaction and free up valuable resources while eliminating burn out on employees. To do so, we had to:

  • Evaluate the state of AI tooling to assist Focal Forensics’ redaction workflows (example: AWS Rekognition and Google Video AI)

  • Determine if AI-assisted software solutions exist in the market or if we needed to build a custom solution for their business (example: Tator)

  • Maintain 100% redaction accuracy

    • Knowing this at the start of the project required us to highlight how AI could assist humans to gain efficiency, but also highlight the need for continuous human correction of AI output.

    • We also needed to adhere to the highest security standards set forth by CJIS and the DOJ

Approach

Our #1 priority was to build trust with the Focal Forensics team. They were searching for years for a way to refine their craft with automation, and were rightly skeptical of AI based on their time and money investment to date. To build and maintain trust quickly we:

  1. Made sure we understood the business and industry before layering on technology

  2. Built a POC in under a week to demonstrate our software development expertise

  3. Arrived ready on day one with deployable software pipeline using AWS Cloud Development Kit

  4. Utilized open source or SaaS solutions where possible

    1. For example, Clerk powers internal user authentication because of a generous free tier and quick time to implement. Since authentication is not a differentiator, why build it from scratch?

  5. Collaborated early and often to remain aligned on a tight timeline

After establishing trust in the first three weeks with a fully functioning end-to-end redaction pipeline and UI, we adopted Linear on the project to manage scope, prioritize, and collect feedback. This allowed us to iterate quickly for the next three weeks and strengthen the custom solution. The team at Focal Forensics appreciated Linear as a way to provide quick and structured feedback, all while keeping the project moving forward in an organized and cohesive manner.  

Starting Architecture:

Crafted: Focal Forensics Starting Architecture

Final Architecture:

Crafted: Focal Forensics Final Architecture

Outcomes

  • Codified Focal Forensics’ standard operating procedure into a custom AI-assisted cloud software solution

    • This saves the team substantial time and eliminates tedious and mentally exhaustive processes that can lead to employee burnout

  • Collected positive beta user feedback and paved a path to full internal adoption

  • Operational efficiencies drive business outcomes and will have a meaningful impact on the Focal Forensics bottom line.

Example of an Un-Redacted Video:

Redacted Video Using Cloud-Based AI Redaction Tool:

What the Focal Forensics Team Had to Say

The talented individuals at Crafted were always responsive to our input and available to discuss any roadblocks or challenges that we came across along the journey. They took the time to learn about our company, our processes, and our needs, which helped us develop a custom software tool that now has the potential to completely change the game for our company.
— Zac Giammarrusco | Co-Founder
Our company did years of research and vetting to find a company like Crafted. Their team, from the very initial meeting, took the time to listen to our specific needs and challenges. Developing software is not something that our company has ever undertaken before. We believed it was essential for us to find a group of individuals who were trustworthy and eager to solve a very difficult problem. We just wish we would have found them sooner!
— Zac Giammarrusco | Co-Founder

Check out their 5-star review on Clutch.

Conclusion

The Focal Forensics team searched for years to find a tool or team to automate their manual redaction process and Crafted was the first team to successfully tackle the challenge. The resulting custom software solution will be a gamechanger for Focal Forensics, and we look forward to watching them dominate the industry. 

In an increasingly competitive and ever-changing AI landscape, Crafted is able to stay lean, make right-sized recommendations quickly, and turn strategy into execution seamlessly. If you are a non-technical company looking to adopt cutting-edge AI tooling, Crafted is your partner. Our expert developers turn AI strategy into execution while remaining laser-focused on value, quality and speed. Plus, as an AWS Partner, we are able to get funding to trial new technology at your company. Reach out today to learn more!

Previous
Previous

Adswerve × Crafted

Next
Next

MedGeo Ventures