Koushal Kumar Email & Phone Number

Gurugram, Haryana, India

At Green Rider Technology, serving the client Outlier, a key entity under Scale AI, I have led and contributed to a range of impactful projects and roles:• In my role as a Prompt Engineer, I trained the Dolphin AI model, achieving a perfect score of 3.0 out of 3.0, significantly improving its accuracy and performance.• As a Quality Lead, I managed a team of 10 members. Our high performance along with other teams was instrumental in securing new AI projects, "Bulba" and "Flamingo."• Developed a robust Flask API to facilitate prompt quality assurance, enabling team leads to efficiently review and approve prompts, ensuring top-notch submissions.• Implemented a "Threads" feature using Django Rest Framework, enabling users to comment on tasks they are assigned to, boosting collaboration and task management efficiency by 30%.• Preprocessed data for a recommendation system algorithm. This algorithm predicts which wholesalers and retailers are most likely to sell a newly launched product based on numerous factors like similar price range, product category, area, margin, etc.

Kochi, Kerala, India

Here, I worked mainly under HP(Hewlett Packard) to do the following:• Analyzed and inspecting different scenarios of occurrence for certain Bugs in Android as well as Windows UI of HP Printers.• Applied Unit testing for Migrating the Servers to LR0 Network to Improve the Cybersecurity.• Collected Data and Introspected the same to determine different size of pages supported for various HP printers• Built Submenu for Diagnosing Sensor Test for SFP Products

Singapore

•Created a IMAGE RETRIEVAL ALGORITHM system using Keras for finding similar images within datasets, akin to Google's search per image feature.•Implemented effective feature extraction techniques, including hand-crafted features and deep learning based approaches like CNNs.•Utilized denoising autoencoders to extract meaningful image features and reduce noise, enhancing image retrieval.•Designed a scoring function to assess image similarity, ensuring relevant images ranked higher using label ranking average precision score