AI Engineering Bootcamp & Internship

Intensive AI/ML bootcamp followed by an ongoing internship at Accenture Baltics, covering model deployment/evaluation, context engineering, and development.

AI Engineering Intern

Mar 2026 – Present

Note: This role involves confidential, NDA-protected work. Specific project names, client details, and proprietary metrics are omitted. Content is described in general terms only.

AI-Powered Developer Assistance Microservice — Team Lead

Led a team of 3 engineers for an internal AI-driven microservice that assists developers by automating contextual analysis of software changes. Organized team meetings, led technical discussions, and authored backlog issues using GitHub Projects to guide development. The system leverages large language models to generate structured insights and delivers actionable feedback directly within the developer workflow — reducing manual effort and improving consistency across engineering teams.

Key Responsibilities

AI Analysis Pipeline: Engineered an AI analysis pipeline that constructs contextual prompts from enriched change data, calls Azure OpenAI models with retry logic, and validates structured JSON responses using Pydantic
RAG Pipeline: Proposed and built a Retrieval-Augmented Generation (RAG) pipeline that indexes Python codebases using AST-based semantic chunking, Azure OpenAI embeddings, and PostgreSQL with pgvector for vector similarity search
Output Publisher: Implemented an intelligent output publisher that updates existing AI-generated comments in-place using deduplication markers, avoiding duplicate notifications to users
LLM Observability: Integrated LangSmith for LLM observability, enabling tracing and monitoring of model inference calls with token usage metadata
Documentation: Authored comprehensive technical documentation including architecture overviews, evaluation reports comparing multiple prompt variants, and contributor onboarding guides

Tech Stack

Python 3.11 FastAPI Azure OpenAI (GPT-4.1-mini, text-embedding-3-small) PostgreSQL pgvector Docker Azure Container Apps GitHub Actions Pydantic LangSmith pytest httpx uv

Skills Demonstrated

LLM Application Development: Integrated Azure OpenAI models into production microservices with structured prompting, response validation, and observability
Async Backend Engineering: Built high-concurrency Python services using FastAPI, async HTTP clients, and event-driven architectures
RAG & Vector Search: Implemented retrieval-augmented generation pipelines with semantic chunking, dense embeddings, and hybrid search (vector + full-text)
Cloud-Native Deployment: Containerized applications and configured CI/CD for Azure cloud infrastructure
Software Quality & Testing: Wrote comprehensive async test suites with HTTP mocking, covering endpoints, pipeline orchestration, retry behavior, and context limits
Technical Documentation: Authored architecture documentation, evaluation reports, and developer onboarding guides

Timeline

AI Engineering Intern

Accenture Baltics

Mar 2026 – Present

AI Engineering Bootcamp

Accenture Baltics

Feb 2026 – Mar 2026

AI Engineering Bootcamp

Feb 2026 – Mar 2026

AI Engineering Bootcamp — Intensive 3-week program covering ML fundamentals, deep learning architectures, and production-grade AI systems through daily lectures and hands-on exercises.

Technologies

Python scikit-learn TensorFlow NumPy Pandas NLTK spaCy Hugging Face Transformers FAISS ChromaDB Pinecone Ollama LangChain CrewAI LangGraph LlamaIndex smolagents sentence-transformers rank_bm25 DeepEval

Key Takeaways

LLM Foundations & Prompt Engineering: Integrated OpenAI Chat Completions API (gpt-4.1-mini/nano) via Azure OpenAI; experimented with model parameters (temperature, max_tokens), prompt variations, and evaluated output quality across clarity, creativity, and response time
NLP & Information Extraction: Built text preprocessing pipelines (tokenization, lemmatization, TF-IDF/BoW); developed regex-based tokenizers handling edge cases (floating-point numbers, negative numbers); implemented fuzzy matching with Levenshtein distance; applied NER using both NLTK and spaCy
Retrieval-Augmented Generation & Search: Built production-grade RAG systems with local LLMs (Ollama/Qwen3 8B/4B), vector databases (ChromaDB, FAISS, Pinecone), and embedding models; implemented BM25 keyword search combined with FAISS semantic search for hybrid retrieval; applied cross-encoder re-ranking (MS-Marco MiniLM) to improve result relevance
Agentic AI & Multi-Agent Orchestration: Implemented ReAct (Reason + Act) agent loops with custom tools; explored agentic workflows across CrewAI, LangGraph, LlamaIndex, and smolagents frameworks covering prompt chaining, routing, parallelization, orchestrator-worker, and evaluator-optimizer patterns; built an AI-powered content moderation system as a capstone challenge
RAG Evaluation & Benchmarking: Evaluated local RAG pipelines using DeepEval metrics (Faithfulness, Answer Relevancy, Contextual Recall/Precision/Relevancy); benchmarked model tradeoffs across speed (tokens/sec), quality ratings, and hallucination resistance; designed test sets with factual, reasoning, comparison, and unanswerable questions to measure retrieval and generation quality