Speech Therapy Platform
Audio AI
Pronunciation Recognition
A custom speech recognition system for children’s therapy, built with phoneme-level modeling, fine-tuned datasets, and production-ready deployment.
Cases
Pardigm
Real-Time Cortisol Tracking App Development for Personalized Stress Management
Key results
- Consistent and accurate LFA recognition across diverse real-world lighting conditions
- Personalized diurnal curves for more meaningful cortisol interpretation
Services- Computer VisioniOS developmentMachine learning
Industry- HealthcareBiotech
TECHNOLOGIES USED- C++SwiftPyTorchCoreMLYolo
Team- 3-5 AI engineers2 iOS developers1 backend developer1 PM1 DevOps1 QA
Duration- 11 months
Location- USA
“We had tried multiple times to solve an extremely complex computer vision challenge in healthcare with no luck. The It-Jim team not only cracked it; they exceeded our expectations. The solution was precise, scalable, and built with deep technical insight. If you're facing a problem that feels nearly impossible, this is the team you want.”
The Challenges & Goals
Pardigm set out to build an AI-powered mobile app that could analyze cortisol levels using LFA test strips – a task that required solving several complex challenges:
- 01
Reliable cortisol measurement through LFA:
Building a CV model to detect and quantify cortisol levels based on subtle color intensity differences on saliva test strips.
- 02
Establishing personalized cortisol baselines:
Enabling the app to learn each user’s natural hormonal pattern and track deviations over time.
- 03
Ensuring model robustness in real-world settings:
Maintaining consistent accuracy despite variable lighting, shadow, and background conditions.
- 04
Handling strip manufacturing variability:
Compensating for minor inconsistencies across different batches of test kits.
Solution Overview
To bring Pardigm's vision to life, we followed a staged development process – from initial proof of concept to real-world testing and development of an AI-powered cortisol tracking app with personalized features. Each phase addressed a specific technical challenge by combining computer vision, AI model development, and mobile optimization. This step-by-step approach ensured the final solution was accurate, user-friendly, and adaptable to real-world conditions.
- Stage 01
Vision and First Steps
When Pardigm first approached It-Jim, their objective was to measure cortisol levels by analyzing the color intensity on salivary LFA test strips. The science behind it was straightforward in theory – but making it work on a smartphone was a different story. Achieving accurate results required custom computer vision algorithms, calibration against variable lighting, and a reliable way to process subtle color differences between test and control zones.
Key initial steps by It-Jim:
- Developed a core algorithm: Using C++, we built an image processing model capable of comparing color intensity between the test and control lines on a neutral white background.
- Enabled quantitative measurements: The system delivered initial cortisol readings in numerical form, providing Pardigm with a crucial proof of concept.
This first milestone demonstrated that real-time cortisol detection via smartphone was technically feasible and laid the groundwork for further development.
- Stage 02
First prototype and real-world testing with Olympic athletes
With the initial version of the app complete, Pardigm moved into field testing. This phase was an important reality check: how would the prototype perform in uncontrolled conditions, outside the lab?
Pardigm arranged testing with Olympic-level runners – a high-stakes environment where precision mattered. Athletes used the mobile app to analyze their salivary cortisol at different times, and the resulting data was shared with our team via Dropbox for evaluation.Outcomes of initial testing:
- Real-world performance feedback: The app successfully processed cortisol readings from test strips used in everyday settings, confirming that our CV model could handle non-lab environments
- Critical insight uncovered: Field use revealed a key limitation – without a personal baseline, single cortisol readings lacked context. To make results meaningful, the app needed to account for natural hormonal fluctuations over time.
This stage validated the model’s foundational accuracy and uncovered the next major requirement: creating a personalized baseline for every user.
- Stage 03
Developing the cortisol baseline measurement
The field testing phase made one thing clear: isolated cortisol readings weren’t enough. To provide actionable insights, the app had to account for each user’s natural cortisol rhythm, which varies significantly during the day.
To solve this, we introduced a baseline tracking feature. Drawing on medical research and feedback from healthcare professionals, we designed a system that prompted users to log their cortisol levels multiple times a day – morning, afternoon, and evening. This data was then used to create a personalized “diurnal curve” that reflected the user’s typical hormonal pattern.Features of the baseline measurement:
- Multiple daily readings: The app prompted users to perform tests at set intervals to establish consistency.
- Personalized curve generation: Each user’s unique cortisol profile was built automatically, capturing daily fluctuations.
- Smarter analysis: Future readings could be compared against this baseline to detect anomalies – elevated levels during rest or unusually low peaks in the morning.
The result was a major leap in user value: instead of isolated numbers, users received context-aware feedback tailored to their own biology.
- Outcome
Final outcome: A prototype ready to transform health monitoring
The project culminated in a fully functional prototype of a mobile app that accurately measured and interpreted cortisol levels in real time directly from salivary LFA test strips.
By combining advanced computer vision, personalized baselines, and real-world testing, the app delivered a unique value proposition: stress tracking that adapts to each user’s biology and works in everyday environments.Key achievements:
- Real-time cortisol insights: Users could get immediate, data-driven feedback to support stress management and recovery.
- Personalized hormonal profiles: The baseline feature made it possible to detect meaningful trends, not just raw numbers.
- Robust performance in real-world settings: The CV model maintained accuracy across varied lighting and background conditions.
- Built-in adaptability: Though initially designed for elite athletes, the system’s flexibility made it equally valuable for other audiences, such as women navigating menopause or individuals facing chronic stress.
This collaboration demonstrated how AI and computer vision can bring lab-level insights into users’ pockets, redefining how personal health can be tracked and understood.
Let’s Talk About Building AI That Works in the Real World
No vague promises. No pressure. Just a real conversation with our technical leadership to understand your needs and explore the best next steps.
What Happens Next?
- Your message goes straight to our core team – not a sales department.
- You’ll talk directly with our CEO, CTO, or a domain expert for a consultation-style first call.
- We’ll ask the right technical questions to understand your challenge and scope.
- If we’re a fit, we’ll follow up with a proposed path forward.
More Projects Like This
Crystal Clear Memories
Computer Vision
3D GenAI
Automated transformation of photos into 3D memorial keepsakes using a hybrid of classical computer vision and 3D generative AI.