Coursera

University of California San Diego

Carnegie Mellon University

University of Southern California

Principal Software Engineer (Deep Learning) at Roche

Saratoga, California, United States

Roche

Principal Software Engineer (Deep Learning)

- Prototyped DNA base-caller from sequencing signals for 4-base sequencing and COVID-19 detection.
-  Built a Hidden Markov Model (HMM) approach for calling pulses on signal data and clustering of pulse types. Developed HMM fitting 3x-8x faster than python’s standard hmm-learn using numba and CUDA.
-  Developed ‘AutoHMM’ parallelization framework to increase throughput, deployable on cloud platforms such as GCP and AWS.
-  Advised statistical methodology to improve biosensor design from weekly experiment datasets.
-  Developed software frameworks to help productionize python code including automated data type control and build regression detection.

Roswell Biotechnologies

Senior Software Engineer / Senior Bioinformatics Scientist

- Helped prototype bioinformatics / data science efforts at a cutting-edge biotech company!
-  Developed cancer event detection algorithms, sample batch effect calibration, statistical machine learning, Bayesian/Frequentist optimization for optical imaging of genomes!
-  Architected stacked ensembles of python notebooks for building and testing bioinformatics / machine learning models.
-  Wrote Python visualization tools for genomic data analysis: Author of EasyPlotly_bio and bed2Ideogram (<a href="https://github.com/prateekt/EasyPlotly">https://github.com/prateekt/EasyPlotly</a>, <a href="https://github.com/prateekt/prattools">https://github.com/prateekt/prattools</a>).

Bionano Genomics, Inc.

Senior Bioinformatics Scientist

- Leading very early stage initiative motivating hybrid systems for sustainable and minimal side-effect negative emissions engineering and saving the world from climate change. Exploring possible approaches (and fusion approaches) using robotics, genomics of climate change, bioengineering, air capture, etc to make a breakthrough in negative emissions engineering.
- Organized fundraisers for National Resources Defense Council (NRDC) on Facebook to deploy solar panels nation-wide and promote conservation efforts.
-  Organized Amazon Shovel Sweepstakes to give free tree-planting shovels to tree planters to plant trees.
-  Advocated for drone tree planting efforts. Served as KickStarter contributor and advisor for Flash Forest, an initiative to plant 1 billion trees with drone technology!
-  Deployed public "Climate Tech" slack (<a href="http://climatetechdiscuss.org">climatetechdiscuss.org</a>) for discussion of existing engineering approaches and education of the audience on their pros/cons. Feel free to join (anyone can join; no technical expertise required!)!

Climate Tech Group

Director, Climate Tech Group

- Worked in the Sebat lab on data science methods for neuro-genomics and applications of machine learning algorithms to bioinformatics problems.
-  Developed pathogenicity predictors to help better understand the functional role of structural variation (and especially non-coding variation) in ASD and SZ.
-  Developed simulator for estimation of mutation rates from large-scale population genetic data.
-  Advised lab members and students on classical machine learning and data mining approaches, advising on mathematical and software development of approaches for biological domain application.
-  Advocated for inclusion and induction of PCR scientists / wet lab experimentalists who analyze zettabytes of data into national Big Data and Computer Science Directorate.
-  Organizer, Machine Learning for Health (ML4H) workshop at NIPS 2017.
-  Posters published at ASHG 2017, ASHG 2018, and SFN 2018.

University of California, San Diego

Postdoctoral Scholar (Psychiatric Genomics)

- Collaborating with neuroimaging clinicians to identify impactful applications of Deep Learning/GPU approaches to medical imaging.
-  Developed approaches to improve analysis of fMRI images for neuroscience of pain with Prof Irina Strigo's lab and UCSD Veterans' Affairs (VA).
-  Applied both classical machine learning approaches (e.g. MDS, SVM, PCA, multi-voxel pattern analysis) and developed newer Deep Learning approaches (e.g. ConvNets, LSTMs) on GPUs in TensorFlow/Caffe,  to improve characterization of signal and noise components in high-dimensional fMRI data.
-  Advocated for engineers/technologists to help aid VA initiatives and improve the health care system.

University of California, San Francisco

Data Science Consultant (Deep Learning for fMRI)

- Organized Quantum Robotics Group (<a href="http://quantumrobotics.org">quantumrobotics.org</a>),  a reading group dedicated to the exploration of quantum computing applications to robotics, bioinformatics, and other domains.
-  Led key lectures on quantum machine learning and its application to quantum robotics.
-  Led group to publish full length book on quantum robotics in Morgan Claypool's quantum computing series (available on Amazon!):  <a href="http://bit.ly/2kM5UDJ">http://bit.ly/2kM5UDJ</a>
-  Won competitive NASA/Google grant to demonstrate quantum algorithms on D-Wave quantum hardware.
-  Founder, <a href="http://quantumdeeplearning.com">quantumdeeplearning.com</a> -- a free and open-source online resource for others to learn about quantum machine learning and quantum robotics. We hope to lower the knowledge barrier to entry for people wishing to contribute to the absolute cutting edge of quantum technology.

Quantum Robotics Group

Director, Founder, Lead Author

- Prototyped low-cost, open-source cyborg backpack and operating software for Android-Linux "Cyborg" Distro (<a href="http://www.cyborgdistro.com">www.cyborgdistro.com</a>). The system allows anyone to have two assistive robot arms, wearable on a backpack. Prototype for shopping market!
-  Prototype controllable by Android phone. Worked on multi-robot-arm control with low-cost EMG sensors.
-  Won "Most Technologically Advanced" at Google Halloween Costume Contest @ CMU 2014, Quarterfinals in Hackaday 2015 Prize Contest, Element 14's 2015 Halloween Contest, and featured by RaspPiPod and Adafruit!

CyborgDistro

Chief Cyborg Officer

- Offer technical advice to startups in the AI/Robotics spaces, helping make crazy engineering ideas feasible. The crazier the idea, the more exciting!
-  Offer expert and broad technical advice in software aspects of many fields of Artificial Intelligence, classical statistical machine learning techniques, and advanced computational data structures and optimization tricks. Have experience with many hardware/software platforms for mobile robots.
-  Regularly field questions on Quora, Facebook, ResearchGate with high credibility rating.
-  Advisor for WiFi Watchdog, an Android application for monitoring elderly indoors using Wi-Fi signals (2012).
-  Advisor for Parking Pigeon, an Android application to help a person find their car in a multi-story parking lot using barometer sensors (2013).
-  Advisor for Mobile Bus Route Tracker Android application for outdoor navigation for persons with disabilities (2013).
-  Advisor for Machine-Learning-based High-Frequency Trading Data Mining Project that applied boosting and estimator fusion to volatility modeling (2012).
-  Advisor for Multi-Robot Debate AI Systems (2014). The vision is to evolve AI systems from the emergent interaction of many debate robotic agents that construct arguments and advance the periphery of knowledge via adversarial debate amongst each other. See algorithmic framework here (<a href="http://tandonp.wordpress.com/2014/05/05/debate-robotics/">http://tandonp.wordpress.com/2014/05/05/debate-robotics/</a>).
-  Advisor for various music software and app development startups (2014).
-  Advisor for HAWK, a machine learning based hate speech detection system for web posts (2016)
-  Advisor for Interpretation and debugging methods for machine learning classifiers in the context of HAWK (2017)
-  Synopsys Science Fair Judge (2015, 2016), Synopsys Science Fair Mentor (2013-2016), Science Olympiad Judge (2014, 2015), Viterbi VSAMP Mentor (2015, 2016, 2017)

TandonRobotics

AI/Robotics Consultant

- Researcher on the Academic Research Initiative (ARI) project to develop robotic and machine learning techniques for mobile radiation sensing in collaboration with Lawrence Livermore National Lab (LLNL).
-  Developed Bayesian Aggregation framework for fusing multiple spatial observations to identify source locations. Algorithm allows simultaneous inference of source type and intensity.
-  Built a Poisson Principal Component Analysis anomaly detector to boost standard PCA methods in flagging suspicious environmental radiation.
-  Led group prototyping of a hardware backpack radiation system controllable via Android phone that analyzes spectrum data in the cloud for threatening sources.
-  Prototyped Active Learning algorithms to facilitate route planning of vehicle-driven sensor for both the single and multi-agent case.
-  Led effort to develop data-analytic calibration scripts for real time sensor hardware system in collaboration with physicists at LLNL.
-  Formulated tractable branch and bound method for detecting a source under both static and dynamic occlusion scenarios, saving lab weeks of sim.
-  Worked on adaptive grid algorithm for multi-resolution testing of large geographic hypothesis spaces of source locations without incurring the computational cost of the finer resolutions. Up to 70% hypothesis space reduction!
-  Led team testing of different localization strategies for use in backpack system. Aided development of localization solutions with Wi-Fi, magnetometer, barometer, and camera-based pose estimation.
-  Advised abstract, modular design/development of Real-Time Bayesian Aggregation (RTBA) and application to medical blood loss domain.
-  Prototyped Zero-Inflated Poisson Match Filter for detecting radioactive sources in pedestrian sensor nets of inexpensive (low-count) wearable spectrometers.
-  Formulated Multi-Sensor PCA and CCA for sensor fault analysis.
-  Built Retrospective Learning for incorporating new fluctuations into a Bayesian hypothesis space.

CMU Auton Lab

PhD Researcher

- Worked on scenario development and visualization tools with Bootstrapped Learning Group.
-  Developed Reactive Action Planning (RAP) algorithms for plane behaviors in flight simulator.

SRI International

Student Associate

- Worked on senior thesis: “A Robotic Controller for Hammering Task” - teaching a robotic arm to use a hammer to drive a nail into a board drawing from motor control studies of human tool use.
-  Read papers on human fMRI and motor control studies on human tool use. Researched applications of human-like control algorithms to robotics.
-  Implemented biologically-inspired neural network controllers for the hammering task based on population coding, reinforcement learning, and inspired by human motor system principles such as Fitt's Law that could learn to use a hammer to efficiently get nails into a board.
-  Prototyped hammering task on Lynxmotion AL5D robot arm.
-  Wrote well-referenced (lots of blog hits!) literature review on controlling robotic arms with biologically-realistic neural networks.

USC Brain Theory Group

Undergraduate Research Assistant

- Researcher on the USC-DARPA Landroids Project to deploy game-theory-based Distributed Constraint Optimization Problem (DCOP) algorithms on robots in a mesh network.
-  Used decision tree analysis to determine suitability of game-theoretic DCOP algorithms for different graph structures to automate switching between algorithms to obtain optimal reward.
-  Implemented DCOP algorithms on a robotic network of 5 iCreates using PlayerStage and lots of custom networking code.
-  Collaborated with USC Wireless Sensing Lab to explore network metrics for optimization by DCOP and mobile robots. Ran experiments on robots to demonstrate benefits of DCOP algorithms on optimizing network Wi-FI signal strengths and data throughput via small motion perturbations.
-  Released DCOP Code as part of the Robotic DCEE Open Source Project (<a href="http://code.google.com/p/roboticdcee/">http://code.google.com/p/roboticdcee/</a>)

USC Viterbi School of Engineering

Undergraduate Research Assistant, Teamcore Research Group

- Prototyped components for DARPA’s CALO project (an early version of SIRI!), applying artificial intelligence techniques to the problem of information overload and personal information management.
-  Built natural language processing algorithms to perform Named Entity Recognition and automate summarization of informal text.

Student Associate in Artificial Intelligence

- Project:Possibility is a nonprofit that builds accessible open-source software for persons with disabilities. Previously served as an Executive Director (2008-2010), helping organize university hackathons in the Los Angeles area. Project:Possibility team won First Prize at the USC Student Innovator Showcase in 2008 and was funded by a Google grant.
-  Regularly maintain several open-source projects as well as network of mentors for programming contests and persons with disabilities. Understood and document needs of users for software projects, while also spearheading ideation for projects and many software development efforts.
-  Participated in team grant cycles and fundraising efforts, helping author technology innovation sections for group.
-  R&D lead engineer and Product Manager of the LunarTabs project, an intelligent accessible guitar tab reader which converts guitar tablature in structured format to screen-reader friendly instructions. Usable with VoiceOver and NVDA, and sports several novel computer-music and human-robot interaction inspired features. Paper presented at CSUN Accessibility Conference 2014.
-  Led organization of first ever USC vs. UCLA SS12, a programming contest to develop software for persons with disabilities in 2010.
-  Built intelligent features for integrated word prediction system in 2008 to help people with limited mobility generate text on a computer. Novel features (at the time): (A) Baseline english language model was generated from wikipedia to ensure maximum predictive capability (B) Learns from a person’s word use so can pick up their language patterns.
-  Worked with team in 2007 to build binary input game playable by someone with limited mobility. Later adapted for use with Neural Impulse Actuator (NIA), a Brain Computer Interface developed by OCZ Technologies.

Project Possibility

Board Member, Executive Director, Robotics Software Engineer

In a former life, I was a jack-of-all-trades software engineer and web developer with extensive programming experience in Java, C++, Java Servlets, PhP/MySQL and some in ASP, Python, and Excel. I advise people on software engineering from time to time but am pursuing a career in Strong Artificial Intelligence now so please do not contact me about software engineering opportunities. Here were some projects:

-  Intern, Network Management Technology Group, Cisco Systems, San Jose. June 2007 – August 2007
-  Vice Chair, Founding Officer, Webmaster. USC Association for Computing Machinery, September 2006 – May 2010.
-  PhP/MySQL Backend Developer, USC Troyspace Project.
-  Website Coordinator, Parkside Area Government. Built Calendar and Newsletter System using Google Calendar APIs, PhP/MySQL, and RSS for Dorm Events. September 2006 – May 2008
-  Excel Data Aggregation Macro and Web Developer for stock trading firm. July 2007.
-  Database Design Consultant for Health Care Database using ASP/Microsoft Access, U.S. Coast Guard. June 2007
-  Web Strategy Consultant for <a href="http://Debatepedia.org">Debatepedia.org</a>. June 2007.
-  Python Developer, Facebook APIs Lead for various web projects. April-July 2007.
- 
-  Team Captain, Monta Vista High School Speech & Debate Club. September 2004 – June 2006
-  Creator, MVLD DB, Evidence Website/Database for Team using java servlets, MySQL, SSL, and more (now would be called a "Web App").
-  Software Tester Intern, JustStudents Web startup. June 2005-July 2005.
-  Programmer Intern for Bibliographic Database Management, Stanford Institute for the Quantitative Study of Society (SIQSS). June 2004 – August 2006
-  Programmer for various AIM/MSN Chat Bot Projects

Prateek Tandon

Prateek Tandon