AI/ML Engineer (Interactive Real-Time Video Avatar with Lip-Sync)
Duration: 0.5 years
Summary:- Built an AI-powered real-time video avatar capable of natural interaction with users, featuring precise lip-sync aligned to generated or user-provided speech
- The system combined face animation models, audio-to-visual synchronization, and dynamic response generation, enabling the avatar to hold live, human-like conversations
- Integrated speech synthesis, real-time rendering, and dialogue management to deliver smooth and responsive user experiences
Responsibilities:- Developed a live animated video avatar supporting real-time interactions with accurate lip-sync;
- Integrated audio-driven facial animation models to map phonemes to mouth shapes and facial expressions;
- Designed and implemented a custom LiveKit plugin to integrate the internal avatar worker into the LiveKit ecosystem and enable seamless interoperability with LiveKit voice agents;
- Built real-time communication pipelines between the avatar worker and LiveKit rooms, including synchronized audio streaming, session orchestration, and avatar state management;
- Built APIs and orchestration logic to manage conversations, avatar states, and user context.
Technologies: Python, FastAPI, PyTorch, OpenCV, Livekit, SyncTalk, InfiniTalk, RunPod, Librosa, aiortc, ElevenLabs, Groq, WebRTC, FFMPEG, Docker, GPU acceleration
AI Developer (AI-Based Buying Committee Detection System)
Duration: 1 year
Summary:- The project aimed to identify the Buying Committee within a company’s contact base using AI
- The system provides a complete toolset to achieve this goal, including PDF parsing and AI-driven analysis to generate detailed company profiles
- It integrates a scraping system that aggregates data from multiple sources via API requests
- The core functionality involves finding the Buying Committee through a custom algorithm built on LLM prompt chains, followed by result validation
- Additionally, the system offers analytics and visualizations for users in the form of interactive charts
Responsibilities:- Designing an AI algorithm based on LLM prompt chains;
- Parsing and validating LLM responses;
- Integrating multiple data scrapers using API requests;
- Building a unified scraping system with contact validation;
- Developing an algorithm for analyzing parsed PDF documents.
Technologies: FastAPI, SQLAlchemy, Alembic, Docker, LangChain, OpenAI, Google GenAI, Pydantic, LangSmith, RapidAPI, Apify, AWS, PyMuPDF, asyncio
AI Engineer (Computer Vision System for Hotel Monitoring)
Duration: 1 year
Summary:- Developed an AI-powered hotel monitoring system based on surveillance cameras
- The system detects people in a video stream, tracks their movements, identifies individuals, and performs re-identification to recognize the same person even when appearing in different camera views
- It also collected and processed data to generate analytics about guest movement across different hotel zones, helping automate monitoring, track visitor flows, and improve space management and hotel security
Responsibilities:- Training OSNet model for person re-identification;
- Collecting and cleaning datasets for training;
- Developing an algorithm based on YOLO, OSNet, and DeepSORT for people monitoring;
- Configuring Docker for real-time streaming;
- Implementing logging and analytics system.
Technologies: OpenCV, PyTorch, Torchreid, OSNet, YOLO, DeepSORT, Logger, Torchvision, FAISS
Back-End/AI Developer (AI Social media automatization)
Duration: 1 year
Summary:- Automatization of clean video creation (full and shorts for social media) and translation using AI
- The user provides a raw video and a text script containing time markers of the chapters, blooped cuts, camera changes, etc
- This video is cleaned automatically
- Video is translated into specified foreign languages using AI speech-to-text and text-to-speech technologies
- As a result, clean videos in all the needed languages are rendered
Responsibilities:- Back-end development;
- Back-end logic for video upload;
- OpenAI integration;
- Prompt engineering;
- ElevenLabs integration;
- DALL-E / Stable Diffusion image generation;
- Testing;
- Error Handling.
Technologies: Python, FastAPI, SQLAlchemy, Pydantic, OpenAI, Anthropic API, ElevenLabs, DALL-E, Stable Diffusion, Whisper, Nginx, Alembic, Redis, Celery, Poetry
Upstaffer since May, 2026
