Tushar T.

Founder and Principal Machine Learning Architect

Raleigh, United States

Experience

Jan 2021 - Present
4 years 11 months
Raleigh, United States

Founder and Principal Machine Learning Architect

Think Tank AI

  • Launched an independent consultancy delivering bespoke AI and machine learning solutions for enterprise clients across healthcare, gaming, and defense sectors.
  • Oversaw full AI product lifecycle from research and data strategy to model development, MLOps infrastructure, and production deployment, emphasizing scalable, cloud-native architectures and robust CI/CD workflows.
  • Designed and deployed an advanced hydration estimation algorithm for TritonX using self-supervised learning, feature engineering, and signal processing, and built a complete MLOps stack with AWS S3/EC2, Snowflake, MLFlow, PostgreSQL, Firestore, Docker, and Kubernetes.
  • Architected machine learning solutions for Zynga including time-series revenue forecasting with Google’s TimesFM, Tableau integrations, Bayesian mixed-media marketing attribution, geo-based pricing optimization, custom NLP pipeline for sentiment analysis, and end-to-end CI/CD pipelines with Databricks and AWS.
  • Led conceptualization and design of an advanced cognitive sensing-based Electronic Warfare system for 0BaseDesign, leveraging transformer architectures, multi-modal data fusion, self-supervised and reinforcement learning to develop explainable, mission-critical tools.
  • Championed AI ethics, responsible data governance, model interpretability, and operational excellence, implementing MLOps best practices with AWS, GCP, Databricks, Snowflake, MLFlow, Docker, and Kubernetes.
Jan 2016 - Dec 2021
6 years
Raleigh, United States

Lead Principal Data Scientist

Valencell

  • Spearheaded development of a cuff-less, non-invasive blood pressure monitoring system by coding and optimizing deep learning algorithms in Keras, TensorFlow, and PyTorch, including transformer architectures.
  • Built the data science team and established a robust MLOps infrastructure from the ground up, designing scalable data management tools for biometrics labs and a hybrid machine learning pipeline.
  • Created a comprehensive CI/CD workflow leveraging Databricks and MLFlow to streamline the end-to-end lifecycle of deep learning models.
  • Presented machine learning value to C-suite executives and partner customers, leading to integration of advanced biometrics (heart rate variability, core temperature, breathing rate) into the product suite, enhancing capabilities and securing critical funding.
Jan 2015 - Dec 2016
2 years
Morrisville, United States

Principal Data Scientist and Research Director

Industrial Heat

  • Led an agile engineering team to validate experimental claims of excess energy generation via fusion reactions.
  • Architected and implemented a scalable MySQL-based data infrastructure for real-time sensor data collection, managing petabytes of experimental data.
  • Developed CI/CD pipelines that automated data processing, model training, and deployment, integrating MLOps best practices for reproducibility and reliability.
  • Designed and integrated a physical sensor system with the database for real-time monitoring, and used OpenFoam heat transfer simulations with OnShape CAD models to validate results.
  • Provided data-driven recommendations to investors by formulating and testing hypotheses through rigorous analysis and replication experiments.
Jan 2014 - Dec 2015
2 years
Morrisville, United States

Principal Data Scientist

Maxpoint Interactive

  • Developed a strategic roadmap for core analytic products, fueling rapid customer growth and positioning the company for IPO.
  • Designed and implemented an advanced analytics engine, leading a data science team in developing, testing, and deploying subsystems for high-impact business outcomes.
  • Applied Bayesian inference and A/B testing to optimize online advertising strategies in real-time bidding environments.
  • Architected and prototyped machine learning and optimization subsystems including Augmented Lagrangian Particle Swarm Optimization (ALPSO) and a cascade PID control algorithm for stable ad bidding.
Jan 2007 - Dec 2014
8 years
Wake Forest, United States

Technical Director of Research and Development

3 Phoenix Inc / Ultra Electronics

  • Led a specialized algorithm team to deliver and deploy advanced production-ready solutions across multiple verticals.
  • Secured contracts in defense, healthcare, finance, product analytics, and biomedical research by leveraging distributed computing frameworks such as Apache Spark and microservices with containerization.
  • Developed multi-modal sensor platforms for DoD, NLP algorithms for fraud detection in healthcare, financial risk models for currency exchange, recommendation engines and Bayesian A/B testing for mobile gaming, and CNN-based video analytics for infant posture monitoring.
  • Spearheaded SBIR grant proposals securing vital research funding and ensured technology transfer to commercial applications.
Jan 2005 - Dec 2007
3 years
Durham, United States

Principal Research Engineer

Signal Innovations Group / Integrian

  • Technical lead for algorithm development team designing video analytics for an adaptive surveillance system.
  • Collaborated with principal investigator to research and document findings in white papers and customer briefings.
  • Developed and integrated a prototype analytic system capable of meta-tagging asymmetric threat events in real-time video streams.
Jan 2004 - Dec 2005
2 years
Durham, United States

Staff Research Engineer

Samsys / Sirit

  • Led algorithm design for the signal processing subsystem of a universal RFID receiver.
  • Created system-level design mapping algorithm processing onto DSP and FPGA hardware subsystems.
  • Implemented a DSP simulation suite in MATLAB and integrated design on ADSP Blackfin processors and Xilinx Virtex 4 using Simulink and System Generator.
Jan 2003 - Dec 2004
2 years
Durham, United States

Staff Research Engineer

MCNC / RTI

  • Led cross-functional team developing a passive vehicle detection and classification system.
  • Supervised analog front-end engineers for sensor signal conditioning and test/integration engineers for field testing of prototypes.
  • Conducted feature extraction with eigenvector analysis of linear predictive coding using ARMA models in MATLAB and Simulink, and developed embedded code on TI TMS320C55.
Jan 2001 - Dec 2003
3 years
Cheswick, United States

Senior Algorithm Design Engineer

Tollgrade Communications

  • Led algorithm development for telephone loop qualification (xDSL/POTS), improving metallic loop impedance measurements via least squares error minimization.
  • Developed time domain reflectometry with spline fitting and neural network backend, and a proprietary frequency domain reflectometry algorithm to qualify loops without metallic access.
  • Implemented algorithms in C and assembly on Analog Devices ADSP-21065L and performed DSP processor comparative studies.
Jan 1999 - Dec 2001
3 years
Santa Clara, United States

Senior Algorithm Engineer

Malleable / PMC-Sierra

  • Led cross-functional team to develop architecture for high-density, low-power voice-over-packet (ATM/IP) processors.
  • Implemented assembly algorithms for speech encoders (G.711, G.726, G.729) and echo cancelers on a VLIW/SIMD network processor.
  • Supervised contractors implementing G.723.1 vocoder and fax relay modules, and collaborated with marketing and executive teams through the product lifecycle.
Jan 1998 - Dec 1999
2 years
Sunnyvale, United States

Algorithm and Firmware Engineer

Silicon Wireless / mDiversity

  • Designed and developed DSP modules for GSM BTS uplink air interface in C and assembly on TI TMS320C542.
  • Implemented patented maximal ratio combining algorithms in a distributed BTS architecture to enhance GSM network coverage.
  • Performed system-level debugging, planned BTS backhaul packetization, and created migration plans for Edge/GPRS.
Jan 1996 - Dec 1998
3 years
Milpitas, United States

Algorithm and Firmware Engineer

C-Cube Microsystems

  • Developed embedded assembly algorithms to decode and display MPEG-1/MPEG-2 bitstreams for DVD applications.
  • Designed error concealment algorithms to maintain signal integrity during data corruption.
  • Created software validation tests using Perl and ClearCase.
Jan 1994 - Present
31 years 11 months

Technical Leadership National Coordinator

Non- BAPS / BAPS Charities

  • Chaired planning committee to create long-term vision and mission, planning grassroots projects nationwide.
  • Developed management structure enabling 100,000 volunteers to deliver humanitarian services.
  • Initiated communication strategy fostering leadership work groups and organized large-scale volunteer events.
Jan 1994 - Dec 1996
3 years
Sunnyvale, United States

Member Technical Staff

Applied Signal Technology

  • Designed and implemented blind estimation, detection, and classification algorithms in TI TMS320C50 assembly for dial-up telephone lines.
  • Developed interface with proprietary TMS OS for memory allocation and multitasking under VxWorks.
  • Created omni-font OCR algorithms for fax banners and multipath channel simulations used by US Army CECOM.
Jan 1992 - Dec 1994
3 years
Berkeley, United States

Member Research Group

Partners for Advanced Transit and Highway Group

  • Modeled outdoor wireless communication channels and LANs for vehicle-to-vehicle communication under PATH.
  • Simulated multiple access protocols (TDMA, CDMA) using Ptolemy and produced feasibility studies for AVCS design.

Summary

I design, build, and deploy analytic products and services. Leveraging over twenty years of diverse technical experience across a wide range of domains enables me to discover unique solutions via broad partnerships across disparate institutions and work groups. I thrive on working in ambiguity to solve complex problems while always looking to innovate by bringing together smart, motivated people and engaging them in technical challenges.

My technical focus has centered on analytics research, machine learning systems, statistical signal processing, sensor systems, statistical optimization, embedded system design and product development. I have deployed products in both commercial and defense industries spanning multiple disciplines - research, software, firmware, and hardware integration. I have a core expertise in rapid prototyping and technology evaluation.

Languages

English
Native

Education

University of California at Berkeley

Master of Science · Electrical Engineering · Berkeley, United States

Rutgers University, College of Engineering

Bachelor of Science · Electrical Engineering · Piscataway, United States

Need a freelancer? Find your match in seconds.
Try FRATCH GPT
More actions