Kylypko
Grygorii

AI & LLM Solutions for Organisations

Design and delivery of production AI/LLM solutions for organisations — from private, self-hosted LLM deployments and retrieval-augmented generation (RAG) over corporate knowledge bases to AI agents that operate internal systems through the Model Context Protocol (MCP). The focus is measurable business outcomes: automated document processing, support and ticket triage, analytics over unstructured data and AI-assisted engineering workflows — with evaluation, guardrails and cost control built in from day one.

• Architected RAG pipelines over internal documentation, tickets, contracts and CRM data — chunking, embeddings, hybrid search (pgvector/Qdrant + BM25), re-ranking and citation-grounded answers
• Deployed private on-premise/VPC LLM stacks (Ollama, vLLM, LM Studio) for organisations with strict data-residency and privacy requirements; model routing between local and cloud models (OpenAI, Anthropic, Gemini) by cost, latency and data sensitivity
• Built AI agents and MCP servers that expose internal APIs, databases and business tools to LLM assistants with role-based access, audit logging and human-in-the-loop approval
• Automated document-heavy workflows — OCR + LLM extraction, classification and validation of invoices, contracts, meter readings and technical reports into structured data
• Delivered LLM-powered analytics over business and IoT data: natural-language-to-SQL, anomaly explanation for energy telemetry, automatic tagging and summarisation of video and text streams
• Established evaluation and quality gates — golden datasets, regression evals, hallucination checks, prompt and model version management, guardrails against PII leaks and prompt injection
• Reduced inference cost and latency through prompt caching, response streaming, batching, model routing and fine-tuning (LoRA) of small models for narrow domain tasks
• Introduced AI-assisted engineering practices (Cursor, Claude Code, Copilot, MCP tooling) and coached teams on adopting them safely

EV Charging Billing Platform (OCPP)

A billing and management platform for electric-vehicle charging networks, built around the OCPP protocol (1.6J and 2.0.1). The system handles charge-point provisioning, real-time charging-session management over WebSockets, energy-metering, tariff-based billing and operator/driver dashboards.

• Designed the OCPP charge-point gateway and the real-time charging-session engine over WebSockets
• Implemented energy-metering and tariff-based billing (kWh, time-based and per-session pricing) with automated invoicing
• Built financial reconciliation and payment-provider integration (Stripe) for driver charging sessions
• Developed operator and driver dashboards with live telemetry, consumption and revenue analytics

Video Intelligence & Stream Processing

Extensive work on large-scale video transcoding and real-time processing and recognition of data from video streams — analysis and automatic tagging of footage from cameras, drones and live streams, plus ad-break detection for ad-supported streaming.

• Designed video transcoding pipelines (FFmpeg) with adaptive-bitrate packaging (HLS/DASH)
• Built real-time stream ingestion and processing (RTSP/RTMP/HLS) from cameras, drones and live streams
• Implemented ML-based recognition, scene analysis and automatic tagging of video content
• Added vision-language models and LLM post-processing for scene descriptions, searchable metadata and automatic summaries of recorded footage
• Developed ad-break detection and cue-point handling (SCTE-35-style markers) for ad-supported streaming

Solar & Energy IoT Monitoring Platform

5+ years of experience building IoT monitoring and automation systems for solar power plants/fields and inverters, tracking energy produced and delivered/exported to the grid. Includes automated monitoring of Ukraine's day-ahead electricity market (РДН) and automated energy sale/export logic, plus energy-metering-node (УЗЕ) monitoring with automatic efficiency calculation, delivered through a custom CRM/admin dashboard.

• Designed IoT data-collection pipelines for solar plants/fields and inverters, aggregating generation and grid-export telemetry in real time
• Implemented automated monitoring of Ukraine's day-ahead electricity market (РДН) and automated decision logic for energy sale/export based on live market pricing
• Built energy-metering-node (UZE) monitoring with automatic efficiency calculation and anomaly detection
• Designed and developed a custom CRM/admin dashboard exposing production, market-pricing and efficiency analytics
• Integrated inverter/device protocols and telemetry ingestion for large distributed sensor networks
• Added an LLM layer over telemetry and market data — plain-language explanations of production anomalies and automatically generated operator reports

NOMO Bank, GB

Digital banking product enabling secure money transfers and payment operations worldwide — spending in-store, online and in apps via Apple devices. My focus was the financial billing and transaction-processing backend behind the mobile application.

• Designed and architected the financial billing and transaction-processing backend, including payment-provider integrations
• Designed database storage and led the migration from DynamoDB to PostgreSQL
• Worked with the product owner to gather, clarify and break down requirements into JIRA tasks
• Code refactoring and performance optimisation
• Drove the Agile process (sprint planning and backlog grooming)
• Led and mentored the sub-team of developers; ran code reviews

PLUTO.TV, USA

Pluto TV, a Paramount (ex. Viacom VCBS) company, is the leading free streaming television service in America, delivering 100+ live and original channels and thousands of on-demand movies in partnership with major TV networks, movie studios, publishers, and digital media companies. Pluto TV is available on all mobile, web and connected TV streaming devices and millions of viewers tune in each month to watch premium news, TV shows, movies, sports, lifestyle, and trending digital series. Headquartered in West Hollywood, Pluto TV has offices in New York, Silicon Valley, Chicago and Berlin.

• Owned content distribution feeds and planned new features with Product Owners across feeds, dashboards and APIs (Apple, Google, Samsung, LG, Vizio, Gracenote and others)
• Built the content-feed admin dashboard and ran partner integrations (Google, Roku, TiVo, Vizio, Samsung, Apple, Gracenote, Comcast, Fire TV) — draft versions, change requests, metadata design and views
• Developed the publishing admin dashboard — a content and asset management system covering all project content, with microservice architecture and a roadmap for front-end dashboards exposing statistics, KPIs and analytics
• Built ad-break detection and cue-point handling (SCTE-35-style markers) for the ad-supported streaming experience
• Worked on video transcoding and processing of live and on-demand content (FFmpeg, adaptive-bitrate HLS/DASH packaging)
• Designed and architected database storage for content (MongoDB + PostgreSQL)
• Worked with the product owner to gather, clarify and break down requirements into JIRA tasks
• Responsible for core component implementation and microservice architecture
• Code refactoring and performance optimisation
• Drove the Agile process (sprint planning and backlog grooming)
• Led and mentored the sub-team of developers; ran code reviews
• Raised integration and unit test coverage to 98% under my technical leadership

H&M, Berlin, Germany

Digital Brand Department. As one of the project, Itsapark - is a multi-brand platform for fashion and lifestyle and we encourage our users to click what they love, read what they want and only buy what they need. We hand-pick better buys and connect people with hot products and brands via inspiring content — from sustainable options and pre-loved pieces to life-long investments. Just the good stuff.

Implementing new Digital Brand initiatives like NFT, Events and Activities in different countries and regions (USA, Japan and Europe), related with Brand promotion and new approaches in Online Shopping Culture.

• Front-end and back-end development, with a primary focus on backend tasks
• Owned the backend of the whole project — microservices, GraphQL API and data exchange between connected stores and 3rd-party APIs
• Responsible for core component implementation and microservice architecture supporting the React UI
• Designed and implemented a large-scale scraping system — collecting product data from marketplaces and running competitive price analysis for the brands on our platform
• Designed and built ETL pipelines and a data warehouse, aggregating multiple sources for analytics and reporting

Astrum LLC, USA

A Machine Learning company focused on quantifying the value of Higher Education. AstrumU’s purpose is to help students find meaningful careers in which they can excel. They’re doing this through leveraging machine learning to drive transparency and efficiencies into the marketplace for the benefit of students, universities, and companies.

Astrum is one of several up-and-coming companies using technology to match employers with recent grads. The company uses machine learning and predictive analytics to help students find impactful careers.

Its end customers are large corporations that are trying to find talented employees.

• Front-end and back-end development, with a primary focus on backend tasks
• Owned the backend of the whole project — microservices, GraphQL API and data exchange between connected services and 3rd-party APIs
• Responsible for core component implementation and microservice architecture supporting the React UI
• Designed and implemented data scraping and ETL pipelines feeding the machine-learning analytics platform

IDEX Exchange

IDEX Exchange (https://idex.market) - IDEX is a decentralized exchange based on Ethereum, featuring ERC20 tokens and Ethereum trading pairs. The platform uses smart contracts that give users the opportunity to manage their private keys, allowing them to trade in a secure peer-to-peer environment. As safer alternatives for fund management, IDEX decentralized exchange features Meta Mask wallet integrations and Ledger Nano S, which are preferable to entering user private keys manually.

IDEX works by leveraging its smart contracts to manage its users’ transactions, unlike other decentralized exchanges. It allows a user to sign off a transaction with their private key but will broadcast the transaction to the Ethereum blockchain itself. In this way, it is able to update account balances and order books in a timely manner. This ensures reduced friction as it allows proper management of the flow of network transactions, providing a centralized exchange user experience without compromising the security and user control of a decentralized exchange.

Part-time engagement as Senior Full-stack Software Engineer — backend refactoring, integration and unit testing, codebase reduction, performance optimisation for block operations within ETH trading, and UI optimisations.

• Backend refactoring and optimisation
• Implementation of integration and unit testing
• Reduced the codebase and optimised performance for block operations within ETH trading
• UI optimisation and refactoring
• CI/CD implementation and support for automated multi-slot deployments

Bosch Global Inc GmbH

Part-time engagement on Bosch Auth Server — a global authorization system for Bosch intranet consumers (Bosch affiliate companies worldwide) covering authorization to internal resources, RBAC, security verification, and notifications across corporate channels (Slack, email, chat, SMS).

• System design, architecture and deployment strategy
• Implementation of the global authorization system for Bosch intranet consumers
• RBAC and security verification
• SSL protection across all auth channels
• Anomaly reporting to external channels for abnormal activity on auth endpoints
• Extended activity logging and alerting on suspicious behaviour
• Notifications and subscriptions across corporate channels (Slack, email, chat, SMS)

Heartland Commerce

Xenial — enterprise restaurant and managed food-service technology, installed in 51,000 locations, hosting 110,000 merchants in the cloud and supporting operations across 62 countries.

Backed by 30 years of experience providing POS systems to restaurants and retail businesses, the company embraced smart mobile devices and cloud computing to modernise its platform.

• Analysis, design and development of user-analytics solutions
• Implementation of bot-detection algorithms
• Front-end and back-end delivery

BotScanner

BotScanner — a browser plugin that determines, in real time, whether each online visitor is a bot, using browser metrics and behavioural signals to detect non-human patterns and rule-block abusive activity that would otherwise overload the server.

• Analysis, design and development of user-analytics solutions
• Implementation of bot-detection algorithms
• Front-end and back-end delivery

SocialBakers

SocialBakers — a global scraping solution that lets companies track social activity across Facebook, Twitter, YouTube and other networks. The platform scrapes metrics for each tracked account (subscriber counts, likes, growth velocity and more) and builds rankings of social pages by country and topic (politics, fashion, travel and so on).

• Built a distributed, cluster-based web-scraping system from scratch (Twitter, Facebook, YouTube, Instagram and other social networks)
• Analysis and classification of scraped data through ETL pipelines
• 3rd-party API integrations
• Database support and system administration