Дима

Dmytro Khaver
Full Stack Engineer
 [відкрити контакти](див. вище в блоці «контактна інформація») q [відкрити контакти](див. вище в блоці «контактна інформація») q https://github.com/VictoriousAttitude
 Ukraine

KEY SKILLS
Languages: Python, Go, Kotlin, Java, SQL;
Backend & Distributed: FastAPI, Django, chi, gRPC, WebSockets, SSE, GraphQL, Celery, Redis, PostgreSQL, event-driven services
AI / LLM: LangChain, LlamaIndex, PyTorch, ONNX Runtime, Claude API, OpenAI API, Whisper, ChromaDB, Qdrant, hybrid search (BM25 +
dense), cross-encoder re-ranking, recall@k, NDCG, MRR, groundedness scoring, golden-set regression, pgvector;
Infrastructure & Ops: Docker, Kubernetes, AWS, CI/CD, canary deploys, Prometheus, Grafana, OpenTelemetry, pprof, SLOs;
Security & Systems: RBAC, audit logging, PII redaction, prompt-injection defense, Android NDK, KMP, NFC/HCE, cryptography, APK
reverse engineering

EXPERIENCE
01/2025 - 08/2026 Senior Backend & AI Engineer (Python/Go)
Netherlands (remote) Interactivated Solutions
• Built a production RAG system for a US-based law firm client, processing ~15 years of archived
legal documents (83K contracts, case briefs, memoranda, and regulatory filings). Implemented
layout-aware PDF parsing with OCR fallback for scanned records, section-level chunking
preserving clause boundaries and cross-references, and metadata filtering by practice area,
jurisdiction, and filing date. (Python, LangChain, Claude API, pgvector, FastAPI, Celery/Redis)
• Deployed the ingestion pipeline with Celery workers processing the full document backlog in
batches with incremental index updates for new filings, role-based access control for 40+
attorneys and paralegals, query audit logging for compliance
• Implemented hybrid retrieval combining dense vector search (sentence-transformers, pgvector)
with BM25 keyword matching and cross-encoder re-ranking: reducing average attorney research
time from 2.5 hours of manual document review to under 45 seconds per query across the full
document archive.
• Wrote the Go query gateway fronting the Python retrieval service: per-attorney session handling,
RBAC across 40+ users, and token-by-token answer streaming over SSE at p95 time-to-first-
token under 900ms. Exposed retrieval to the gateway over gRPC with generated stubs and
contract tests in CI. (Go, chi, gRPC)

08/2023 - 12/2024 Senior Backend Developer (Python/Go)
Riga, Latvia (remote) Nitra
• Built the Go model-serving gateway fronting ONNX Runtime workers: adaptive request batching,
health-checked routing, and circuit breaking on worker failure. Sustained 3.2K RPS at p99 under
45ms across 3 nodes, replacing a Python serving layer that required 8. Instrumented with the
Prometheus Go client behind Grafana SLO dashboards. (Go, chi, FastAPI, ONNX Runtime)
• Built the tool-calling agent behind shelf compliance: Claude function schemas over catalog
lookup, planogram diff, and image re-check tools, with a bounded execution loop, per-tool
timeouts, and structured output validation that retried malformed calls against a repair prompt.
Collapsed multi-step operator queries from three round trips to one. (Python, Claude API tool
use, Pydantic)
• Built the ML model pipeline (PyTorch -> ONNX -> ONNX Runtime Mobile) for on-device image
recognition on Android, reducing inference latency. Managed model versioning and OTA updates
via custom FastAPI endpoint.
• Built the Go model-serving gateway fronting ONNX Runtime workers: adaptive request batching,
health-checked routing, and circuit breaking on worker failure. Sustained 3.2K RPS at p99 under
45ms across 3 nodes, replacing a Python serving layer that required 8.
Instrumented with the Prometheus Go client behind Grafana SLO dashboards. (Go, chi, FastAPI,
ONNX Runtime)
• Performed security analysis and APK reverse engineering using jadx and apktool.
• Full responsibility of Certificate Security Analysis, APK Signing Security and Security Auditing.
• Developed a Python-based document processing service (LangChain, Claude API, pgvector) for
automated extraction and semantic search across NFC protocol specifications, APDU command
references, and payment scheme technical documents.
• Cut LLM spend 71% with a two-tier semantic cache keyed on embedding similarity and
confidence-based model routing that escalated to the larger model only on low-scoring
retrievals, holding answer quality flat against the evaluation set. (Redis, Claude API)
• Integrated vector embeddings (sentence-transformers) into KMP shared modules for cross-
platform semantic product search, replacing keyword-based SKU lookup with AI-powered fuzzy
matching.
• Wrote the real-time NFC relay in Go: goroutine-per-session APDU forwarding over WebSockets
with buffered-channel backpressure, context-scoped cancellation on teardown, and errgroup
supervision of reader and writer pairs. Held 4K concurrent sessions per node at p99 relay latency
under 80ms; pprof heap profiling exposed a per-session buffer leak whose fix cut steady-state
memory 40%.
• Architected the end-to-end NFC payment relay across 5 protocols (NfcA, NfcB, NfcF, NfcV,
IsoDep) with Host Card Emulation, biometric authentication, and certificate security analysis.
EXPERIENCE
09/2021 - 05/2023 Senior Full stack Developer (Python/Go)
Tel Aviv, Israel (remote) Voicenter
• Built the WebRTC signaling gateway in Go: goroutine-per-session WebSocket handling for call
setup, SDP offer and answer relay, ICE candidate exchange, and lifecycle tracking with graceful
reconnect. Sustained 6K concurrent signaling sessions per node at p99 setup latency under
120ms. (Go, gorilla/websocket)
• Wrote the Go telemetry ingestion service consuming WebRTC getStats from live sessions,
computing MOS via the ITU-T G.107 E-model with streaming aggregation of jitter, packet loss,
and RTT, exported to Prometheus and alerted in Grafana.
• Integrated Google Accessibility Checks into the CI pipeline, automating WCAG contrast, touch
target, and content description validation on every UI test run.
• Ensured all call status indicators met WCAG 1.4.1 (no color-only signalling) and 1.4.3 contrast
ratios, verified with Accessibility Scanner across all app themes.
• Wrote Python automation scripts for VoIP call quality analysis — parsing WebRTC stats logs,
aggregating jitter/packet-loss metrics, and generating weekly reports for the QA team.
• Built internal CLI tooling (Python, Click) for automated APK build distribution and Slack release
notifications, reducing manual release overhead.
• Contributed to the backend call-routing service (Python, FastAPI) — implemented webhook
handlers for call event logging and real-time status sync with the Android client.

01/2020 - 12/2021 Middle Python Developer
England (remote) Artificial Intelligence Castings Ltd
• Built and maintained the ML pipeline for headshot analysis and talent-to-role matching using
computer vision (InsightFace, CLIP)
• Integrated voice profiling for accent/tone classification (Resemblyzer)
• Assisted with data ingestion, model serving and deployment on AWS (Docker, Celery, Redis)
• Developed and maintained RESTful APIs (FastAPI) for talent search, profile ingestion, and casting
brief matching, serving internal and client-facing endpoints.
• Wrote integration tests (pytest) for model endpoints and set up monitoring for inference latency
and embedding drift in production.
• Integrated speech-to-text (Whisper) for automated audition transcription, feeding structured
dialogue data into the matching pipeline.

01/2017 - 12/2019 Junior Full Stack Developer (Android/Backend)
Los Angeles, US (remote) Bizico
• Worked in a cross-platform mobile team. Co-developed an Android and IOS applications,
enabling businesses to showcase alcoholic beverages in partner locations.
• Contributed to Android application development with the integration of beacon technology and
participated in intense testing period Bluetooth LE for automated user check-in/out functionality.
• Created dynamic UI that adapted in real-time based on user's location and proximity to Bluetooth
devices, significantly enhancing user experience and app engagement.
• Participated in Android Fingerprint API development for smooth client's authentication
• Joined back end (python) team for the big outsource client: designed and shipped GraphQL
resolvers in Django, migrated legacy REST endpoints to FastAPI, wrote async task workers
(Celery + Redis) for report generation and batch data processing.
• Implemented the reporting pipeline aggregating BLE proximity events into business-facing
analytics.

EDUCATION
09/2017 - 02/2019 Master's degree
Computer Software Engineering Kryviy Rih Economic Institute

09/2013 - 06/2017 Bachelor's degree
Computer Software Engineering Kryviy Rih National University

LANGUAGES
English Fluent Russian Native Ukrainian Native