DEEP_LEARNING • NLP • COMPUTER_VISION • GENERATIVE_AI • AGENTIC_AI • DEEP_LEARNING • NLP • COMPUTER_VISION • GENERATIVE_AI • AGENTIC_AI •
ABOUT
_ME.exeProgramming and frameworks
PythonC++HTMLCSSJavaScriptTypeScriptSQLPyTorchTensorflowFastAPINumpyPandasMatplotlibFlaskLangChainOpenCVMongoDBscikit-learnSQLite
Tools and platforms
GitGithubLinux ShellPostmanJupyter NotebookAWS EC2AWS S3HuggingFace
Core competencies
Data ScienceArtificial IntelligenceNatural Language ProcessingComputer VisionGenerative AIPrompt EngineeringQualitative ResearchData VisualizationResearch WritingData StructuresAgentic AI
Soft Skills
AdaptabilityProject ManagementCross-functional CollaborationInterpersonal communicationProblem-SolvingEvent OrganizationTeam Leadership
WORK_History
AI Intern
Onsuriy Technologies
August 2025 - Current | Banglore
Mission_Logs:
- [1]Boosted health-report accuracy by 28.6% through optimized agentic LLM workflows and advanced reasoning prompts.
- [2]Built a multilingual transcription pipeline with 100% timestamp alignment, enabling CSAT analysis across 6+ languages.
- [3]Engineered 58+ LLM and rule-based fraud-detection triggers, increasing high-risk claim identification accuracy.
- [4]Developed a scalable FastAPI architecture reducing trigger-execution latency by 40% and enabling seamless frontend integration.
- [5]Reduced manual claim-review workload by 50% via consolidated fraud-scoring logic and automated signal aggregation.
Tech_Stack:
LLMsFastAPIPythonAgentic AIFraud Detection
PROJECTS
- >Architected a scalable, Retrieval-Augmented Generation (RAG) pipeline to drive enterprise-grade knowledge retrieval.
- >Engineered a dynamic multi-persona system, optimizing information delivery across disparate department workflows.
- >Delivered a modular solution capable of switching knowledge bases effectively, enhancing cross-functional utility.
RAGPythonLLMs
- >Conducted comprehensive benchmarking of computer vision frameworks (MediaPipe vs. YOLO V7), driving the strategic selection of YOLO V7 for superior real-time performance.
- >Deployed a high-throughput, low-latency pose estimation model capable of analyzing complex workout movements (e.g., Shoulder Press) with precision.
PythonYOLO V7StreamlitMediaPipe
- >Pioneered a hybrid Deep Learning architecture integrating Spatial Pooling with LSTMs to unlock silent speech recognition capabilities.
- >Optimized feature extraction pipelines to process navigational and phonemic cues from video data with high fidelity.
Deep LearningLSTMSpatial Pooling
RESEARCH
& PUBLICATIONSHow could we add emotional nuances to AI-generated music?
February 2025 - June 2025Journal Paper
ISLR using Deep Learning: Attention is Everywhere
September 2024 - December 2024Accepted
AI in Finance: Navigating Ethical Quandaries
August 2024 - September 2024Book Chapter
TROPHIES
Certificate of Achievement
IIIT Delhi
Fork-it Hackathon (2nd Position)
Certificate of Achievement
Bennett University
Engineering Project Showcase 2023 (3rd Position)
Vice-President
Bennett University
Artificial Intelligence Society (AI club of BU)
AI Expert
Bennett University
GDSC BU (Google Developer Student Club)
Tech Head
Bennett University
ICosmic'22
GET IN TOUCH
Contact Information
Send a Message
Made with ❤️ by Moaksh Kakkar
© 2026 All right reserved.