MIKHAIL GORELKIN

Principal AI Scientist Mathematician Engineer

email: ad1df4@r.postjobfree.com

phone: 972-***-****

blog: http://blog.mikhail.gorelkin.com

summary Principal AI Scientist, Mathematician, and Engineer proficient in taming business uncertainty and complexity through artificial intelligence and generative AI, advanced data science, multi- agent and mathematical modeling, and engineering in production. key skills Artificial Intelligence, NLP, Machine Learning & Deep Learning, Reinforcement Learning, Probabilistic Programming, Decision Making

Generative AI: custom AI chatbots & autonomous AI agents using ChatGPT API and open- source Large Language Models, fine-tuning open-source LLMs, integrating them with data and computations, etc.

Advanced Methods in Data Science: transformers and zero-shot transfer learning for big data forecasting and probabilistic predictions in complex systems, quantum-inspired optimizations, recommender systems, etc.

Complex Systems & Multi-Agent Modeling

Mathematical Modeling

Python, C++, C#, Scala, Julia, SQL, scikit-learn, TensorFlow / Keras, PyTorch, spaCy, Hugging Face, LLMs / Transformers, LangChain, Faiss, Pinecone, Weka, PySpark, Akka, Google Colab, Amazon SageMaker, Amazon EC2 / Linux, GitHub Copilot. experience AI & Advanced Data Science Consultancy (Boston, MA) 04.2005 – current Principal AI Scientist Mathematician Engineer

Areas of work

Research using scientific publications

Data analysis and conceptual design of innovative approaches, models, and algorithms

Development of models and algorithms (prototyping and programming for production)

Development of distributed and multi-agent systems, including agent-based modeling Projects (sample list)

Generative AI / LLMs (BridgeCare Technologies Advisory): Researched recent advances in publications and open source. Designed a conceptual framework. Prototyping.

Big data forecasting (Celsius Holdings): Analyzed big data and developed a probabilistic solution for inventory forecasting (PyMC 5, PySpark, Amazon SageMaker).

Intelligent automation (Deloitte / ServiceNow): Researched and developed for production symmetric & asymmetric semantic search on multiple NLP tasks (PyTorch, spaCy, Transformers, Faiss, Dataiku).

AI for complex systems (AES Corp.): Researched and prototyped: several probabilistic forecasting models for gas prices (Prophet, Edward / TensorFlow), a framework for full simulation on the Southern California energy market (multi-agents, Bayesian Nets), a latent reinforcement learning algorithm, and a multi-agent reinforcement learning for competitive games model for the Colombian hydro energy market (PyTorch). Developed for production an automatic gas trader for the Southern California energy market.

Nonparametric probabilistic inference (Scientific Systems Company, Inc. / DARPA): Researched and developed a model for nonparametric inference for simulated trajectories with noise based on a variational autoencoder and probabilistic classifier (TensorFlow Probability).

Intelligent advertising (ADΣXT Corp.): Analyzed data streams from different platforms and restored their integrity with algorithmic constraints. Developed a reinforcement learning algorithm for optimal advertising strategies on FB (TensorFlow). Designed and prototyped a probabilistic multi-agent test framework for simulation and analysis of the users' behavior on FB.

An intelligent virtual agent (agent.ai): Researched automatically generated responses for IVAs. Developed a model and all algorithms based on deep adversarial learning (Torch7 / Lua).

Scalability for an enterprise system (Earley Information Science / Ford): Analyzed, solved, and developed a solution for a scalability problem for Ford's Virtual Assistant for Vehicles project.

Clinical trials (BBCR Consulting): Designed and developed an adaptive computational framework for CTs (Julia).

NLP / NLU (noHold): Developed several NLP / NLU algorithms for production with spaCy and scikit-learn. Prototyped unsupervised deep embedding for text clustering (word2vec, MXNet) and several semantic transformations on the texts based on the dependency parser (spaCy).

Intelligent mortgage (Boston Consulting Group): Researched and designed a framework for mortgage loan automation.

An intelligent online dating platform (Likeli): Researched and designed a conceptual model for the platform. Designed a personality similarity metric between users and implemented in Theano / GPU library. Developed an intelligent multi-agent framework based on Scala, Breeze, Akka, and Redis to scale computations for many thousands simultaneous users in real-time. Developed computing personality profiles for FB users based on the Big Five Personality Traits Model. Developed a real-time personalization system based on users’ feedbacks for several types of bandit algorithms modified for a vector space: Epsilon- Greedy, Softmax, UCB1, Bayesian, and Bayes-UCB1. Improved Bayes-UCB1 for balance multi- factor learning. Solved scalability and performance issues for 1M users and put the system into production.

Marketing (MutualMind): Developed for production various algorithms such as word sense disambiguation, Twitter hashtag decomposition (used Norvig’s algorithm), topic modeling, sentiment analysis, etc. (Python, NLTK, Gensim).

A marketing platform (SocialExtract): Developed many NLP algorithms to clean, transform, and integrate data streams from different sources. Prototyped a library of NLP and ML primitives and a framework for the quick composition of computational pipelines based on Python functional and streaming libraries. Prototyped a relevance engine in this framework. After the client was satisfied with the accuracy, I programmed it in a distributed environment (Hadoop / MapReduce, MongoDB, and Redis) for scalable production with mrjob / Python and Java. Implemented the distributed Sparse Proximal SVM based on publications.

A recommender (IEEE): Researched novelty detection in scientific articles and Deep NLP as a framework for NLP problems. Designed and developed a multi-label classification for a large label set (> 7K) (Mulan, MEKA, scikit-learn) and a similarity metric as a core for the client's recommender system.

Intelligent real-time logistics (Weft): Researched and designed a conceptual framework for a real-time logistics platform based on DARPA’s Cougaar architecture. Developed several heuristics based on ant-colony optimization for adaptive routing optimization (Java).

A social network (PopCircle): Researched, designed, and developed for production a composable approach to collaborative filtering with several social regularizations (Octave).

E-learning (Junyo): Researched and designed a model for adaptive learning. Developed the concept map extraction from documents and a fuzzy matching algorithm (Java, Stanford NLP). Found significant patterns in the data for sales prediction (Weka). Compuware Corp., Technology Department (Detroit, MI) 03.2000 – 11.2004 Software Developer VII

Researched and developed many advanced features for QALoad (automated performance and scalability testing software) (MS C++/STL, all major RDBs, Statistica). Central Transport International, Inc. (Sterling Heights, MI) 03.1998 - 02.2000 Systems Architect

Managed a team of ten engineers to developed an NT-based distributed enterprise architecture for 100+ terminals across the US, Canada, and Mexico using satellite communication. Managed several consulting teams to implement an ERP system on clusters.

Resource Technologies (Troy, MI) 08.1997 – 03.1998 Software Consultant

Developed a scalable architecture for terminal operations based on MTS and MS SQL Server with up to 300 MS-DOS clients / hand-held computers using MS RPC (MS C++). After several months of work, the client hired me as a Systems Architect. Advanced System & Designs, Inc. (Troy, MI) 01.1996 – 05.1997 Software Engineer

Researched and developed the DoES for Windows (the primary product for the Shainin approach to Statistical Design of Experiment) and the ANOVA-TM 2.x for Windows (the primary product for the Taguchi approach to DoE) (MS C++, Statistica). DataNet Technologies, Inc. (Troy, MI) 02.1993 – 06.1994 Software Engineer

Researched and developed all statistics for the WinSPC (statistical process control software), including the non-normal capability analysis for all types of Pearson’s and Johnson’s distributions (MS C++).

education Voronezh State University (Voronezh, USSR) Master of Science, Mathematics

Nonlinear functional analysis. The master's thesis: Diffeomorphisms in Banach spaces. Continuing Education with Certification

Train & Fine-Tune LLMs for Production (Intel Disruptor Initiative) 2023

Quantum Computation using Qiskit v0.2X (IBM) 2022

Quantum Computing (Coursera, Saint Petersburg State University) 2021

Algorithmic Information Dynamics (Santa Fe Institute) 2018

Parallel Programming (Courser, École Polytechnique Fédérale de Lausanne) 2017

Text Mining and Analytics (Coursera, University of Illinois at Urbana-Champaign) 2015

Statistical Learning (Stanford University) 2014

Mining Massive Datasets (Coursera, Stanford University) 2014

Decision Making in a Complex & Uncertain World (University of Groningen) 2014

Data Mining with Weka. Parts 1 & 2 (University of Waikato) 2014

Introduction to Dynamical Systems and Chaos (Santa Fe Institute) 2014

Introduction to Complexity (Santa Fe Institute) 2013

Game Theory (Coursera, Stanford University) 2013

Natural Language Processing (Coursera, Columbia University) 2013

Algorithms: Design and Analysis. Parts 1 & 2 (Coursera, Stanford University) 2013

Machine Learning (Coursera, Stanford University) 2012

Model Thinking (Coursera, University of Michigan – Ann Arbor) 2012

Contact this candidate