Applied AI Engineer · London, UK

Chuck Bolevics

I design, build, and ship AI-powered products end-to-end for early-stage teams.

RAG systems, LLM applications, ML pipelines, and full-stack product work — from prototypes to production-facing internal tools.

Work

Selected projects

A selection of applied AI and data work across regulated, high-stakes environments.

Natural Language → SQL Analytics Platform

Finance

An AI agent that lets analysts query complex financial databases simply by typing questions in English.

Unlocked over £10M in follow-on funding and is scaling to millions of customer workflows, turning hours of SQL work into instant insights.

Speech-to-Insight Pipeline

Defense

An AI pipeline that transforms raw audio into structured intelligence using diarization, topic modeling, and automated summarization.

Boosted transcription throughput by 20%, helping teams process mission-critical recordings faster and more accurately.

AWS → Snowflake ETL Fabric

Finance

An automated data ingestion fabric that moves, validates, and structures data from AWS into Snowflake.

Provided analytics teams with governed, query-ready datasets—reducing manual data wrangling and accelerating decision-making.

Trustworthy Vision Stack

Defense

A full computer-vision pipeline with accuracy monitoring, data lineage, and built-in bias controls.

Achieved 95% detection accuracy while meeting strict governance requirements for high-stakes defense environments.

LLM Guardrails + Evaluation Framework

Finance

A real-time governance layer for enterprise LLMs, featuring automatic scoring, alerts, and safe fallback responses.

Enabled compliant, auditable LLM usage—reducing operational risk and accelerating adoption of generative AI in regulated organizations.

Defense Report RAG System

Defense

A retrieval-augmented system that searches large internal corpora and surfaces the most relevant intelligence instantly.

Replaced week-long manual report hunts with seconds-fast results, radically improving analyst productivity.

Independent work

Things I build outside client work

ParagonAI

AI / OSS

An open-source RAG workspace that builds a custom knowledge base from user-provided URLs, extracting and indexing relevant content.

Enables natural-language querying over large, curated corpora—returning precise, context-aware answers.

View on GitHub

Splash Language

Languages

A Haskell-built domain-specific language for reading integer columns from files and applying composable arithmetic operations.

Removes boilerplate by letting users express numeric transformations concisely.

View on GitHub

CookPal AI

Health · Coming soon

An AI meal-prep planner that converts fitness goals into tailored menus, macros, and grocery lists.

Automates meal planning by generating personalized nutrition-aligned prep plans with no manual work.

Experience

IBM UK

London, UK · Jun 2021 — Present

May 2023 – Present

Software Engineer, Machine Learning

  • Developed an NL2SQL agent-driven interface for PostgreSQL using Vue.js, Tailwind, and Python; the POC enabled natural language workflows for a financial client and unlocked GBP 10M+ in follow-on funding.
  • Partnered with cross-functional squads to deliver scalable RAG systems (Milvus, LLaMA, Sentence Transformers) so IBM clients can retrieve high-signal internal knowledge quickly.
  • Engineered a governance framework with real-time guardrails by integrating Ragas, watsonx.governance, and Nvidia TensorRT-LLM to monitor production LLM behavior.

Jun 2022 – May 2023

Machine Learning Engineer

  • Designed and deployed computer-vision object detection for the defense sector using PyTorch and TensorFlow, trained on 14K+ images and optimized for edge inference.
  • Built a speech-to-text NLP pipeline (fine-tuned Whisper, diarization, keywording, summarization) on AWS with BERT models, cutting analysis time by 20%.
  • Produced and demoed ETL proof-of-concept flows on Snowflake and AWS S3 that helped secure a major financial-sector contract.

Jun 2021 – Sep 2021

Software Engineer Intern

  • Built a full-stack 3D digital twin for client factories with Python, Vue.js, Flask, Babylon.js, and IBM Object Storage to help customers trace products in a circular economy.
  • Created a prototype digital card for Google Wallet and Apple Pay (Java + Swift) that stored recycling-generated vouchers to encourage sustainable buyer behavior.
  • Led a four-person team through five agile sprints with weekly IBM mentor reviews and client workshops, delivering a client-approved MVP.