With Cerebellum Capital, InFoCusp engineers tackle the problem of financial forecasting by making models that continuously auto-upgrade to estimate market movements by analyzing non-stationary time series data. This information is then utilized into making automated strategies for portfolio management.
In collaboration with Cerebellum Capital technical team, InFoCusp engineers have devised a fault-tolerant data flow system that can fetch, process, validate and analyze data coming from disparate sources with many complex dependencies.
InFoCusp engineers work with LegalSifter team on Natural Language Processing and Machine Learning algorithms for "ContractSifter" (Website). ContractSifter extracts useful information from the legal contracts, automatically categorizes the information into clauses and other legal fields (such as identifying dates of the contract, finding termination clause statements, liability clause statements), and presents the information to the end users in a lucid manner. The product also highlights non-standard terms, conditions and legal clauses to the end user of the product, based on the learnings from the automated analysis of their legal corpus.
RunScribe devices (Website) are one of the most advanced running wearables that are commercially available. A RunScribe device captures thousands of data points every single run. The device and the data-analysis platform integrated with the device provides athletes, trainers and runners an easy-to-understand and meaningful summary about their run data, enabling them to make better training decisions. InFoCusp engineers work with RunScribe team on their real-time data-driven analytics platform, and also help the team derive better running metrics.
InFocusp personnels helped Jawbone CTO team on heart rate estimation algorithms for bio-impedance sensors providing support for sensors validation and signal quality analysis, noise cancellation from signals, devising low-power and low-memory versions of the models. We worked with the team right from the design phase till the commercial launch of the Up3 product. (Website).
Lot of exploratory research work was done to estimate blood glucose from BodyMedia sensor sets. We collaborated with world's best researchers and doctors to make better non-invasive glucose estimation techniques. Some of the work is presented in medical and chemical engineering journals (Paper1) and (Paper2) as an example.
BodyMedia had also designed a clinically validated arm-based ECG sensor. With BodyMedia, InFoCusp worked on heart rate algorithms from these arm-based ECG signals. The challenging part was to reduce the EMG noise from the arm-ECG signals.
With BodyMedia, InFoCusp team designed accurate energy expenditure for their wearable devices. Using the raw sensor data (accelerometers, heat flux, Galvanic Skin Response), both the teams combined to make extremely accurate algorithms for estimating energy expenditure (calorie tracking). This was done by utilizing sophisticated sensor fusion and machine learning techniques. This work was recognized as an impactful innovation in industrial AI, and received an award in IAAI-2011. The work was also featured in the AI magazine (Link to article). To accurately solve the energy expenditure problem, the procedure relied on a few intermediate algorithms and estimation models : such as activity classification, steps estimation, sleep duration and sleep quality estimation, body posture estimation.
The energy expenditure estimation algorithms continuously evolved, and InFoCusp researchers and engineers have been instrumental for designing, validating and deploying those algorithms into various firmware platforms between 2009-2015 (until the BodyMedia products were phased out after its acquisition by Jawbone), and hundreds of thousands of customers used these clinically validated products to be healthier and fitter.
The work of better energy expenditure algorithms continued after BodyMedia's acquisition - and InFoCusp team carried on working with Jawbone engineers.
InFoCusp worked with Vancive (an Avery Dennison Company for medical technologies), to bring the BodyMedia physiological modeling algorithms suite to their bio-adhesive wearable sensor technology named “Metria (TM)”. We also helped the team develop the software user interface to communicate with the Metria devices.
For a local client (startup in a stealth mode), we helped on algorithmic aspects of an online advertising platform product. Our work contribution was to design and algorithm to cater relevant ads to end-users, while also attempting to maximize the revenue of the advertising platform by improving the user experience by enhancing their click through rate and lead generation.
For a confidential client (a US hospital), InFoCusp had devised and engineered a monitoring system based solely on visual feedback. The camera mounted on a patient's room would provide feed to the vision-based algorithms. These algorithms authenticate doctors and nurses based on their unique codes on the uniform, and collect statistics automatically about their visits (number of visits per day, duration of the visits) and their adherence to the hygiene protocol (did they wash hands in the sink area, did they wear gloves or not while treating the patient).