Chemist · AI engineer · scientific systems builder

I build AI that can survive scientific scrutiny.

I’m Lars, a registered chemist and AI developer. I build evidence-grounded systems for chemistry literature, spectral data, document extraction, and research workflows—keeping validation and human judgment inside the loop.

Explore selected work GitHub LinkedIn

Flagship work: Synth Assistant · Agentic Chemometrician
Current: AI Developer · technical lead on selected projects
Focus: Scientific AI · document intelligence · chemometrics

BS Chemistry, Magna Cum Laude · MS Chemistry University of the Philippines Diliman

Document Intelligence◆Agentic Workflows◆Chemometrics◆Scientific Computing◆LLM Evaluation◆Structured Extraction◆

Flagship work

Two systems at the center of my scientific AI work.

Both projects treat scientific reliability as a product requirement: structured evidence, deterministic tools, explicit validation, reproducible artifacts, and human review before consequential conclusions.

01 · Flagship

Chemistry literature · experiments · reproducibility

Synth Assistant

An evidence-grounded research system that turns chemistry papers into structured records, connects literature procedures to experimental results, and prepares traceable outputs for further analysis and manuscript development.

Extracts synthesis parameters, characterization data, author rationale, and source-linked summaries
Supports experiment comparison, procedure ranking, FTIR processing, and proposed next-experiment workflows
Uses immutable runs, schemas, annotation metrics, prompt/model versioning, and reproducibility artifacts

PythonPydanticPyMuPDFSQLiteFTIRLLMs

Explore Synth Assistant

02 · Flagship

MCP · spectral analysis · human-in-the-loop AI

Agentic Chemometrician

An MCP-based scientific analysis prototype where an AI agent orchestrates bounded, deterministic chemometric tools while preserving human approval as the gatekeeper for scientific claims.

Guides dataset inspection, planning, preprocessing, modeling, validation, interpretation, and report generation
Surfaces leakage, imbalance, small-sample, split-stability, and suspicious-metric risks
Distinguishes the best-measured model from the best-defensible model and records failures, fallbacks, and approvals

PythonMCPscikit-learnNIRFTIRHuman review

Explore Agentic Chemometrician

Supporting work

Experiments, tools, and reusable engineering systems

Document AI03

LLM OCR Experiments

Benchmarks OCR and vision-language parsing strategies for production document-extraction pipelines, with ground truths, scoring, cost comparisons, and failure analysis.

OCREvaluationClaudeOpenAI

Repository ↗

Quantum chemistry04

ORCA Reviewer

A reproducible ORCA 6.1 learning repository connecting theory, input design, output checks, spectroscopy, metal complexes, and antioxidant thermochemistry.

ORCADFTTD-DFT

Repository ↗

Research agents05

Research Report Writer

An agent-assisted research application for source discovery, evidence gathering, web content processing, and structured report generation.

AgentsStreamlitResearch

Repository ↗

Data engineering06

Excel / CSV Cleaning

Automates multi-sheet splitting, formula refresh, AI-assisted boundary detection, pandas cleanup, and standardized Excel/CSV export.

pandasOpenAIxlwings

Repository ↗

Agentic search07

Deep Search with Agno

A multi-agent research workflow with adviser and researcher roles, scientific-paper search, general web tools, and API-facing execution paths.

AgnoFastAPISearch

Repository ↗

AI engineering system08

My AI Boilerplate

A reusable library of coding skills, review workflows, architecture guidance, testing practices, security checks, and agent-orchestration configurations.

CodexClaudeSkillsAgents

Repository ↗

About

A scientist’s instinct for evidence, applied to software.

I began in coordination and bioinorganic chemistry, synthesizing new ligands and studying metal-ion interactions through spectroscopy and crystallography. That training made provenance, uncertainty, controls, and reproducibility feel non-negotiable.

Now I build LLM pipelines, extraction systems, agents, and machine-learning workflows. Synth Assistant and Agentic Chemometrician are where those two backgrounds meet most directly: one organizes chemical evidence and experiments; the other makes spectral modeling more structured, reviewable, and defensible.

Ground the output

Keep evidence, source locations, schemas, and uncertainty attached to generated results.

Evaluate the system

Use golden sets, failure analysis, cost and latency benchmarks, and regression tests.

Respect the expert

Automate routine work while preserving review gates for consequential decisions.

Experience

From laboratory teaching to production AI.

2025 — present

AI Developer Offshorly

Builds LLM pipelines, structured extraction services, RAG systems, and agent workflows. Leads technical planning and delivery on selected projects.

2025

AI Engineer DOST–Advanced Science and Technology Institute

Worked on government AI initiatives involving RAG agents, routing, data ingestion, and GPU-enabled NLP systems.

2023 — 2025

Instructor IV UP Diliman Institute of Chemistry

Taught chemistry laboratories, scientific writing, data analysis, and instrumental methods including UV–Vis, FTIR, pXRD, and TGA.

2023 — 2026

MS Chemistry Researcher University of the Philippines Diliman

Developed new aminoquinoline-derived ligands for optical Cu(II) detection, combining synthesis, spectroscopy, crystallography, and computational chemistry.

Capabilities

A cross-disciplinary toolkit.

Enough chemistry to challenge the model. Enough engineering to build the system around it.

Applied AI & LLM systems +

Agent orchestration, MCP, RAG, structured outputs, document intelligence, prompt and context engineering, evaluations, guardrails, model routing.

Software & data engineering +

Python, FastAPI, Flask, Pydantic, pandas, PostgreSQL, Docker, AWS, Azure Service Bus, testing, observability, API design.

Scientific computing +

Chemometrics, scikit-learn, spectral preprocessing, PCA, SVM, model interpretation, ORCA, DFT and TD-DFT workflows.

Laboratory chemistry +

Coordination chemistry, synthesis, UV–Vis, fluorescence, FTIR, pXRD, scXRD interpretation, TGA, experimental design, scientific writing.

Let’s build something useful

Scientific AI, chemistry research, or a stubborn document pipeline?

I’m open to research collaborations, AI engineering conversations, and PhD opportunities beginning in 2027.

simondlarslenon@gmail.com Connect on LinkedIn ↗