Олександр

Oleksandr Bunin
Senior Full-Stack Engineer | Python, Java, React
[відкрити контакти](див. вище в блоці «контактна інформація») | [відкрити контакти](див. вище в блоці «контактна інформація») | Kyiv | [відкрити контакти](див. вище в блоці «контактна інформація»)

Summary
Senior Full-Stack Engineer with 8+ years building production-grade, high-concurrency SaaS, specializing in Python
with FastAPI, Java with Spring Boot, and React, backed by PostgreSQL and Redis and deployed on AWS and
SAP BTP with Docker, Kubernetes, and CI/CD. Builds AI agents in Python with LangChain and LangGraph
integrated with SAP Joule, SAP's AI copilot, through FastAPI services, and built an AI workspace orchestrating 40+
LLM, image, and video models with RAG and SSE streaming. Architected enterprise platforms serving 500+ tenants
and tens of thousands of users, led a legacy to microservices to Spring Modulith evolution, and ran an observability
and resiliency program that cut MTTR roughly 60%. Applied Mathematics background, comfortable owning the full
path from data model to frontend experience.

Technical Skills
Python & Backend: Python, FastAPI, Pydantic, asyncio, SQLAlchemy, Celery, pytest, Java, Spring Boot, Spring
Modulith, Spring Security, Hibernate, Golang, gRPC, REST
Frontend: React, Next.js, TypeScript, TanStack Query, Zustand, Redux Toolkit, Tailwind CSS, shadcn/ui, Reusable
Component Libraries
ML & Data: NumPy, pandas, scikit-learn, PyTorch, Hugging Face Transformers, Embeddings, Vector Search,
pgvector, Model Serving, Jupyter
AI & Agentic: LangChain, LangGraph, SAP Joule, RAG, Multi-Agent Orchestration, Tool Execution, LLM
Orchestration, Prompt Engineering, Context Engineering, SSE Streaming
Data & Messaging: PostgreSQL, Redis, MongoDB, MySQL, SAP HANA Cloud, Kafka, RabbitMQ, SAP Event
Mesh, Protocol Buffers
Cloud & DevOps: AWS, Docker, Kubernetes, Helm, Terraform, CI/CD Pipelines, GitHub Actions, Jenkins, SAP
BTP, Cloud Foundry, Cloudflare
Architecture: Microservices, Event-Driven Systems, DDD, CQRS, SAGA, Strangler Fig, Multi-Tenant SaaS,
Resilience Patterns

Professional Experience
06/2024 – Present Senior Full-Stack Engineer
Remote LeverX International LLC
High-performance internal SaaS used daily by SAP Account Executive teams worldwide to
model enterprise deals across TCV, ACV, LEP, and IT1 and IT2 calculations over multi-
year, multi-product scenarios.
Integrated the platform with SAP Joule, SAP's AI copilot, building Python services
with LangChain and LangGraph agents so account executives can query and work
with deal data in natural language.
Developed FastAPI services exposing deal and pricing capabilities as AI-callable tools for
Joule, with Pydantic-validated contracts and async request handling.
Led the platform's evolution from a legacy monolith to microservices to Spring
Modulith, using the Strangler Fig pattern to migrate incrementally and cutting p95
latency by replacing HTTP and Event Mesh calls with in-process events.
Designed an immutable snapshot store for every deal change, serialized with Protocol
Buffers and made storage-efficient through structural sharing, paired with a domain
event log for audit and analytics.
Built real-time multi-user collaboration over WebSockets with SAP Event Mesh and
optimistic concurrency control, surfacing edit conflicts to clients to prevent lost updates.
Added Resilience4j Circuit Breakers and idempotent Retry with backoff to stop
cascading failures when upstream SAP systems degraded.
Cut MTTR by roughly 60% over a year through Jaeger tracing, structured ELK
logging, and customer-impact-tuned alerting.
Led the React and TypeScript UI, building a reusable component system with
optimistic updates and rendering tuned for multi-thousand-line deals.

Enforced DDD aggregates and domain invariants, with architecture tests guarding
module boundaries at compile and test time.
Redis for caching of frequently accessed deal and pricing data, reducing repeated
database reads and keeping scenario calculations fast under load.

03/2024 – 06/2024 Senior Software Engineer, 3-Month Contract
Remote Promtheon
AI SaaS platform aggregating 40+ LLM, image, and video models into a single agentic
workspace.
Led the backend data layer and APIs in Python with FastAPI for chat, document
processing, and agent tooling, using async endpoints and Pydantic schemas, with
Redis-backed caching and dedup keys that cut redundant model calls and controlled
cost under repeated context.
Built multi-step AI agent workflows with LangChain and LangGraph for autonomous
reasoning and tool execution, managing context windows, tool-call routing, and
intermediate state to keep long agentic loops coherent and cost-controlled at scale.
Built RAG pipelines for document processing, covering chunking, embeddings, and
vector retrieval, so agents answer from user-supplied documents with grounded context.
Engineered low-latency SSE streaming for agent responses, handling backpressure and
mid-stream cancellation to keep time-to-first-token low under concurrent sessions.
Led the Golang orchestration layer routing requests across 40+ LLM, image, and video
providers behind a unified API, so new models onboard via config alone.
Drove end-to-end delivery on AWS with EC2, RDS PostgreSQL, S3, and CloudFront,
behind Cloudflare with Docker-based deploys, defining the service patterns,
conventions, and observability for the platform.

04/2018 – 03/2024 Senior Backend Engineer
Remote LeverX International LLC
Multi-tenant SaaS on AWS dispatching work to tens of thousands of field technicians
across 500+ customer tenants.
Improved the dynamic plugin compilation pipeline, cutting compilation latency by
roughly 90% through response compression and caching of compiled artifacts, removing
a recurring source of slow plugin loads under multi-tenant load.
Developed features across a 20+ microservice ecosystem on Java and Spring Boot plus
selected Node.js integration services, while maintaining strict multi-tenant domain logic
and tenant isolation across the platform.
Built Kafka-based event flows using the Transactional Outbox pattern to guarantee
database and event consistency, eliminating dual-write inconsistencies that previously
caused divergent state between services in the dispatching pipeline.
Integrated Jaeger distributed tracing across 20+ services, turning cross-service debugging
from hours of log-grepping into minutes.
Optimized PostgreSQL performance under production load by restructuring queries
and indexing to reduce query execution time and raise system throughput during peak
dispatching volume.
Wrote backend unit and integration tests plus Python E2E suites with pytest,
sustaining roughly 90% coverage with E2E runs gating production deployments on
Kubernetes and Helm.

Education
09/2024 – 02/2026 Master's Degree, Computer Software Engineering
Open International University of Human Development "Ukraine"
09/2016 – 06/2020 Bachelor of Science in Applied Mathematics
Kyiv Polytechnic Institute, Physical-Technical Institute