AI Engineer · Python & C · Data & Systems
Engineer with 6+ years driving data analysis, process automation, and predictive maintenance at scale. Now channeling that analytical foundation into AI/ML — building LLM pipelines, RAG systems, and data-driven tools with Python and C.
Open to roles in AI/ML Engineering, Backend / Data Engineering and Software Development.
RAG Against the Machine - Hybrid Search & Repository-Scale RAG
An end-to-end RAG system designed to interact with large codebases, combining sparse and dense retrieval to provide highly accurate, natural-language answers using the Qwen LLM.
- Semantic Code Chunking: Intelligent, context-aware chunking strategies to preserve code logic across multiple files.
- Hybrid Search Architecture: Fuses BM25 lexical scoring with ChromaDB vector embeddings via Reciprocal Rank Fusion (RRF).
- Evaluation Pipeline: Built-in Recall@k evaluation suite to benchmark and optimize retrieval quality.
- End-to-End CLI: Robust command-line interface managing everything from data ingestion to answer generation.
Call Me Maybe — LLM Constrained Decoding
Deterministic LLM pipeline that guarantees 100% valid JSON function calls by masking model-generated logits at the token level. Enforces strict type-matching against JSON Schema and custom stopping criteria during generation — zero retries, zero parsing failures.
Jobjective — AI Job Market Analyzer
Web application that scrapes real job postings, validates and deduplicates data via Pydantic models, synonym dictionaries and English normalization, then generates a personalized skill-gap dashboard. Live at jobjective.net and on Play Store (Closed Beta).
Fly-In — Multi-Drone Routing Optimizer
Simulation engine that solves optimal multi-drone routing using Time-Expanded Graphs. Handles pathfinding, collision avoidance, and capacity constraints — implemented from scratch without external graph libraries.
Pac-Man — Arcade Game Engine
Full recreation of the classic arcade game: MVC architecture, procedural maze generation, ghost AI with distinct behavioral personalities, and persistent highscores.
Codexion — Deadlock-Free Threading Simulation
Low-level concurrency simulation in C with configurable scheduling algorithms. Focuses on mutex management, race condition prevention, and resource ordering to guarantee deadlock-free execution.
