Загруженный файл CV

TECHNICAL SKILLS

• Languages
  Python, TypeScript, JavaScript, SQL,
  C#, Java, C/C++

• AI / ML
  RAG, hybrid dense+sparse retrieval,
  Qdrant, embeddings (OpenAI,
  SPLADE), reranking (Cohere), LLM
  integration (Claude, GPT), FastAPI
  model serving

• Frontend
  React, Next.js, three.js / web-ifc
  (WebGL), D3.js

• Backend & Data
  Node.js, REST APIs, PostgreSQL +
  Prisma, Redis, Clerk auth, web
  scraping

• Cloud & DevOps
  Docker, nginx, DigitalOcean, RunPod,
  Linux, Git

• Security
  Web-app hardening, upload validation,
  SSRF mitigation, rate limiting

SUMMARY

Full-stack developer and Systems Engineering graduate (Politehnica University of
Timişoara) who designs and ships production Al products end to end. Sole architect
of a live RAG application over Romanian construction law
spanning a React/Next.js frontend, a hybrid retrieval pipeline over a vector
database, a Python ML inference service, and full cloud deployment. Strong in
Python and TypeScript, security-aware (Hack The Box CWES), with a habit of
owning problems across the whole stack.

FLAGSHIP PROJECT

Application

• Production RAG / AI Web
  Founder & Sole Developer
  Built and shipped a production full-stack Al application solo, owning the full
  lifecycle from architecture to live public beta.
  Engineered a hybrid RAG pipeline: dense (OpenAI) + sparse (SPLADE) search
  in Qdrant, fused via RRF and reranked with a Cohere cross-encoder; answers
  cite the exact source document and page.

• Added LLM query understanding and streamed answer generation with Claude
  (Sonnet/Opus) over SSE, including a deep-reasoning mode; built a separate
  FastAPI microservice serving the SPLADE model.

• Implemented an in-browser BIM/IFC 3D viewer (web-ifc WASM + three.js)
  feeding a semantic model summary into the Al context; developed a companion
  Revit add-in in C# with auth handoff to the web app.

• Added a Redis response cache, role-based collaborative projects
  (owner/editor/viewer), and an interactive D3.js knowledge graph.

• Hardened the app: PDF upload validation (MIME + magic-byte sniffing), SSRF
  protection on user-supplied URLs, and request rate limiting.

• Integrated Clerk, Neon serverless PostgreSQL (Prisma), DigitalOcean and
  RunPod GPU compute into one product; deployed with Docker + nginx.

• Built a document-ingestion pipeline (PDF → Markdown → vector database) for
  the knowledge base.

• Wrote Python tooling to orchestrate RunPod GPU resources for embedding/AI
  workloads; developed web scrapers that collect and structure external data
  into PostgreSQL.

• Built a Telegram bot automating file ingestion/conversion and managing RunPod
  and DigitalOcean Spaces - chaining multiple APIs and cloud services into
  one workflow.

CERTIFICATIONS

• CWES― Certified Web Exploitation
  Specialist, Hack The Box
  ID: HTBCERT-84AE33F4F4

EDUCATION

• B.Sc. in Systems Engineering
  2022 - 2026
  Faculty of Automation & Computer Engineering, Politehnica University of
  Timişoara (UPT)

• High School Diploma
  2010 - 2022
  "Ioan Vodǎ" Theoretical Lyceum, Cahul, Moldova

LANGUAGES

• Romanian native
• Russian - native
• English - professional

ACHIEVEMENTS

• 2nd place at ITEC (ACI Worldwide-sponsored hackathon).
• Finalist, UniVs Threats 2025.