Anubhuti Bhardwaj_

●​ Conducted R&D and built a POC for automated medical test kit analysis using computer vision. ●​ Trained a custom YOLOv11 object detection model and integrated OCR and color theory, implementing concurrent request execution for scalable inference via FastAPI endpoints. ●​ Attained a 70–85% reduction in end-to-end latency, improving processing time from ~110 sec to ~35 sec. ●​ Designed and deployed a real-time user & event recommendation system using profile metadata, geolocation, and interest signals. ●​ Developed a FastAPI-based backend to serve low-latency recommendations and integrated PineconeDB for vector similarity search using embedded user and event metadata. ●​ Improved recommendation accuracy by 90%, enabling more relevant user-event matching based on profile. ●​ Built an agentic AI chatbot on local LLMs to deliver context-aware and mood-sensitive conversations. ●​ Designed a Python microservices-based agentic pipeline with a master agent orchestrating specialized sub-agents, using a vector database for content retrieval and Ollama for private, on-device inference. ●​ Achieved GPU-based inference with 2–3 s end-to-end response time, enabling privacy-preserving. ●​ Built a 3D virtual try-on platform allowing users to visualize clothing on a personalized 3D avatar and simulate tailoring adjustments in real time. ●​ Developed microservices using Blender to generate 3D avatars. Integrated open-source 2D and 3D models and created a pipeline with queue-based processing to generate try-on results and simulate tailoring changes.

At Technoculture, I developed microservices with Dapr, FastAPI, and Pydantic, integrated with TursoDB, applying clean coding practices, modular abstractions, and efficient low- and high-level code management. I also built a personal knowledge graph library using ontologies, enabling efficient creation, querying, and management of complex graphs, and implemented vector-based similarity search support (VliteVSS, SQLiteVSS) with LiteLLM + instructor frameworks for content-grounded and persistent reasoning.

• Developed and maintained scalable RESTful APIs using Django Rest Framework, integrating OpenAI APIs to power intelligent automation. • Enhanced video and audio processing workflows by integrating ffmpeg, reducing processing time by 45% and improving video output quality and reliability. • Developed and published Python automation packages, 100% reused by dependent services, cutting repetitive development tasks and reducing implementation time for new features. • Conducted load and performance testing using Locust, identifying bottlenecks and achieving 25% higher throughput under peak load. • Implemented asynchronous task queues using Redis and Celery, enabling non-blocking video processing and reducing average task latency by 40% while improving system responsiveness. • Managed bug resolution and error handling across APIs using FFmpeg, reducing response time by 45%, while delivering comprehensive API documentation to enable smooth cross-team collaboration.