eg. Emad Gohari

My Work

Building the feature.
And what supports it.

My work spans the visible AI feature and the less visible foundations: context, evaluation, data, and repeatable experiments. Here’s a closer look at what I’ve built and contributed to.

Download resume ATS-friendly PDF

Selected projects

A closer look at the work.

Three examples across production AI, research infrastructure, and agent tooling.

LLM applications

Grounded summaries for complex business reports

Thomson Reuters 2024 Production

My contribution

I built the LLM summarization workflow for CLEAR Business AI: selective XML extraction, targeted calls for individual report sections, and source links for summary claims. I integrated search and LLM verification to help distinguish entities with the same name, then refined extraction and prompts through subject matter expert feedback.

  1. Select context

    Relevant XML fields and report sections

  2. Summarize

    Targeted LLM calls for each section

  3. Ground claims

    Source links and entity verification

Explore the technical details

The engineering challenge was deciding which context each call needed, keeping claims traceable, and incorporating expert feedback into the next iteration.

Built the LLM summarization workflow for CLEAR Business entity reports, live in production processing hundreds of reports daily. Designed selective XML extraction for targeted per-section context covering business overview, personnel, adverse media, and legal records. Integrated Bing Search and LLM verification to reduce same-name entity mismatches and linked claims to report sources. Refined extraction and prompts through SME annotation rounds against exemplars for 10-15 companies, and compared GPT-4 and Claude Sonnet on quality, latency, and cost to inform model selection.

Tools & technologies: Python · GPT-4 · Claude Sonnet · Bing Search API · Prompt & Context Engineering · LLM Evaluation · XML Processing

ML infrastructure

Making research experiments repeatable

Thomson Reuters 2026 Internal Tool

My contribution

I designed, implemented, and tested a SageMaker Pipelines retraining workflow for CLEAR’s entity-resolution research. Researchers’ feature-generation, training, and evaluation code became reusable DAG steps, with parallel experiments tracked in MLflow. Alongside that internal workflow, I built Snowflake proofs of concept to assess migration feasibility.

  1. Prepare features

    Reusable research processing steps

  2. Train & evaluate

    On-demand SageMaker workflow

  3. Track experiments

    Parallel runs recorded in MLflow

Explore the technical details

The retraining workflow was an internal research tool. The Snowflake work was proof of concept. The time saving reflects an estimate of researcher effort per experiment round.

Designed and built an on-demand model-retraining pipeline on SageMaker Pipelines, turning researchers' feature-generation, training, and evaluation code into reusable DAG steps. Enabled parallel experiments tracked in MLflow, cutting experiment turnaround by an estimated ~40%. Built proof-of-concept Snowflake processing and training pipelines and assessed the gaps for migrating AWS research workflows to Snowflake.

Tools & technologies: Python · AWS SageMaker Pipelines · MLflow · PySpark · Snowflake · XGBoost · Poetry

Agent tooling & evaluation

Making investigation agents easier to configure

Thomson Reuters 2025 - 2026 Production

My contribution

I contributed to CLEAR Investigate, Thomson Reuters’ commercially launched agentic investigation product. My work focused on a PydanticAI configuration layer for system prompts, tool definitions, and agent/subagent wiring. I also migrated existing LLM-as-judge evaluation tooling into a reusable repository.

  1. Configure prompts

    Central system-prompt definitions

  2. Connect tools

    Tool definitions and agent wiring

  3. Support evaluation

    Reusable LLM-as-judge tooling

Explore the technical details

Configuration brought prompts, tool definitions, and agent connections together so the team could change them consistently. The evaluation work organized an existing LLM-as-judge capability into a repository the team could reuse.

Contributed to CLEAR Investigate, Thomson Reuters' commercially launched agentic AI product for investigations. Built a central PydanticAI configuration layer for system prompts, tool descriptions and availability, and agent/subagent wiring, enabling experimenters to reconfigure agents from one place instead of editing scattered definitions. Migrated the team's existing LLM-as-judge evaluation capability into an organized repository for reuse and future CI/CD integration.

Tools & technologies: Python · PydanticAI · Claude · Agent Harness Engineering · Multi-Agent Orchestration · LLM-as-Judge Evaluation · Prompt Engineering

Experience

The broader picture.

Eight years working across research, product features, and the systems that connect them.

Thomson Reuters

Jul 2021 - Present

Toronto, Canada

Senior Machine Learning Engineer II (Research)

Applied ML at TR Labs for CLEAR, Thomson Reuters' due-diligence platform for KYC/AML screening and investigations, and Checkpoint, its tax research platform. Built LLM features and agent workflows, search models, training pipelines, and secure data infrastructure with researchers, subject-matter experts, and product engineering. Used SQL/PostgreSQL for entity-resolution data analysis and name/address candidate blocking before downstream ML matching and ranking.

Work projects

Entity Resolution Retraining & Snowflake Pipelines: CLEAR 2026 Internal Tool Read story Agentic Investigation System: CLEAR Investigate 2025 - 2026 Production Read story CLEAR Business AI: GenAI Report Summarization 2024 Production Read story
Entity Resolution Data Infrastructure: CLEAR KYC/AML 2022 - 2024 Internal Tool

Built research data infrastructure for CLEAR's entity-resolution system spanning ~800M entities and billions of documents, with a 700 idents/second target. Designed a PII-isolated AWS environment with SageMaker Studio, CloudFormation, and per-researcher profiles; defined a versioned schema reconciling two incompatible annotation sources, with Spark and Python pipelines to merge records and surface label conflicts; and replaced manual runs with scheduled, repeatable SageMaker Pipelines.

Tools & technologies: Python · Apache Spark (PySpark) · AWS SageMaker · AWS CloudFormation

INAGO Inc.

Jul 2018 - Jun 2021

Toronto, Canada

NLP & Machine Learning Engineer

Built NLP and language-understanding systems end to end, from data preparation and model training to evaluation and application integration, including transformer fine-tuning and research collaborations with university and industry partners.

Work projects

Automated Question Generation from Documents 2020 Research

Fine-tuned the open-weights T5 model to generate questions from documents, cutting manual data-curation effort by 40%. Experimented with input representations and evaluated generated text with BLEURT in a research collaboration with university partners.

Tools & technologies: Python · PyTorch · Hugging Face Transformers · T5 · BLEURT

Domain-Specific Language Understanding 2019 Research

Trained domain-specific Word2Vec embeddings and LSTM-based NLU models, with interpretability testing to make model decisions transparent.

Tools & technologies: Python · PyTorch · Word2Vec · LSTM

Conversational Question Answering with NLP & LSTM Models 2018 Research

Developed a conversational question-answering system combining syntactic and semantic document analysis with LSTM-based text tagging. Used semi-automatically generated ontologies to interpret user questions and identify intent, and experimented with knowledge graphs for knowledge representation in an industry research collaboration.

Tools & technologies: Python · spaCy · LSTM · Text Tagging · Semi-Automatic Ontology Generation · Knowledge Graphs

Technical expertise

Capabilities, with context.

The tools matter most when they support a useful result.

LLMs & Agents

Grounded generation, prompt and tool configuration, and evaluation informed by domain experts.

Agentic Frameworks (PydanticAI) · Prompt & Context Engineering · LLM Evaluation (LLM-as-Judge, SME Review) · OpenAI API

More tools & skills

LLM Agents & Tool Use · Agent Harness Engineering · Multi-Agent Orchestration · Grounding & Source Citation · Model Selection (Quality, Latency, Cost) · GPT-4 · Claude

See summarization work

Machine Learning & NLP

Language understanding, sentence embeddings, and search that reflects the intent of a question.

PyTorch · Hugging Face Transformers · Sentence Embeddings · Semantic Search

More tools & skills

Machine Learning · Deep Learning · NLP · Transformer Fine-Tuning (T5) · Model Training & Evaluation (F1 Score, ROC-AUC, MRR, MAP, Recall@k) · Text Classification & Text Tagging · Clustering Models · Information Retrieval · Knowledge Extraction & Representation (Knowledge Graphs, Ontologies) · LSTM · scikit-learn · XGBoost · spaCy · NLTK · pandas · NumPy

See semantic search work

MLOps, Data & Cloud

Reusable training workflows and secure data foundations that help researchers run better experiments.

AWS SageMaker (Studio, Pipelines) · Apache Spark (PySpark) · Experiment Tracking (MLflow) · Snowflake

More tools & skills

MLOps · ML Pipelines · Data Pipelines · AWS (S3, EMR) · AWS CloudFormation · PostgreSQL · Elasticsearch · OpenSearch

See retraining work

Languages & Practices

Research code turned into tested, reproducible components, in close collaboration with researchers and product engineers.

Python · Java · Scala · Bash

More tools & skills

SQL · Agile Development

See infrastructure work

Education

Foundations.

M.Sc. in Computer Science

York University

2014 - 2018 · Toronto, Canada

Thesis: Interactive Question Answering Using Frame-based Knowledge Representation

GPA: 8.17 / 9

B.Sc. in Computer Engineering

Amirkabir University of Technology

2010 - 2014 · Tehran, Iran

GPA: 17.18 / 20

Publications

Research & writing.

  1. Question-worthy Sentence Selection for Question Generation

    Canadian AI 2020 · 2020 · co-authored

  2. Interactive Question Answering Using Frame-based Knowledge Representation

    York University M.Sc. Thesis · 2018

  3. Time Aware Topic-based Recommender System

    Big Data & Information Analytics · 2016 · co-authored

  4. A Study on Prediction of User's Tendency Toward Purchases in Websites based on Behavior Models

    Information and Knowledge Technology (IKT), 6th Conference, IEEE · 2014 · co-authored

Get in touch

Have a problem worth
working through?

I’m interested in useful AI applications, thoughtful engineering, and teams that care about getting the details right.

Let’s talk Find me on LinkedIn