# Subham Kumar Das > AI/ML & SWE(5+ YOE) | MSc @ Brown(23-25) | Founding ML Engineer | Open Source Contributor | Ex- SDE II @ Reliance Jio & HighRadius Profile: https://flows.cv/subhamkumardas I’m an AI/ML engineer who has worked on agentic AI workflows using LangGraph, production RAG systems with LangChain, Pinecone, and custom embeddings, and real-time deep learning pipelines for perception and signal analysis. I graduated in 2025 with a Master’s in Computer Science from Brown University, focused on AI and systems, and bring 5+ years of experience building production-grade ML systems alongside backend platforms. My background also includes microservices, Kafka, SQL, cloud infrastructure, and scalable APIs, which gives me a strong foundation for taking AI systems from experimentation to reliable deployment ## Work Experience ### Founding ML Engineer @ Type 1 Compute Jan 2025 | United States • Built and deployed real-time deep learning pipelines for event-based perception and signal analysis, implementing projects such as SpikeYOLOv8 Tracker and Spectrum Analyzer using CNN, RNN, and Transformer inspired architectures for low-latency, on-device inference in resource-constrained environments. • Developed and extended the TALON SDK to support end-to-end model workflows, including training integration, graph transformation, simulation, profiling, hardware-aware partitioning, and deployment, enabling faster experimentation and productionization of advanced ML models. • Driving development of neuroscience-inspired AI models optimized for low-power, on-device inference (sub-1 W, sub-millisecond latency) to enable real-time adaptation & decision-making on edge devices like low-earth orbit satellites. • Implementing and refining high-speed object-tracking solutions using event-based cameras (e.g., Prophesee EVK4), converting asynchronous visual data into actionable input for neuromorphic processors. • Designing sensor-fusion pipelines and embedding learning loops so devices continuously adapt to new stimuli and environments without any cloud dependence. • Collaborating with hardware and systems teams to align ML architecture with brain-inspired processors, ensuring models meet edge constraints (power, latency, memory). • Presenting findings and progress to stakeholders (engineering leadership, partners, investors) and helping guide the strategic AI & compute roadmap from model design to deployment. ### Gen AI Engineer @ Norric AI Jan 2025 – Jan 2025 | United States • Architected a production RAG system using LangChain, Pinecone, and a custom embedding model, building end-to-end retrieval workflows that included document preprocessing, chunking, metadata-aware indexing, query rewriting, hybrid search & reranking, to improve relevance and robustness for ambiguous & multi-hop queries. ### Founding Software Engineer @ ReplyQuickAI Jan 2025 – Jan 2025 | United States • Developed and optimized the in-house voice pipeline, building a flexible microservice architecture to reduce dependency on external APIs, lowering costs & providing a fallback solution. • Contributed to building and improving platform features using tools like Supabase, Vercel, Twilio, and RetellAI. • Ensured real-time performance by logging response times, costs, and error rates for benchmarking and reliability. •Contributed extensively to the Messaging Service, fixing critical issues in opt-in/opt-out flows, state synchronization, and message delivery logic. • Improved template rendering and personalization by synchronizing user data across Supabase and Prisma, ensuring correct contact resolution and message history. • Enhanced follow-up messaging logic, ensuring correct handling of automated message triggers, quiet hours, and unsubscribed states. • Identified and resolved multiple bug and edge cases through systematic feature testing and integration testing of N2 modules. ### Machine Learning Engineer (RA), Eric Morrow Lab, MCB @ Brown University Jan 2024 – Jan 2025 | Providence, Rhode Island, United States Deployed a custom MoSeq pipeline based on Auto-Regressive Hidden Markov Models (AR-HMM) to process 3D depth capture from Xbox Kinect, segmenting behavior motifs to support longitudinal phenotyping and assess drug intervention outcomes in rodent models for Christianson Syndrome. Leveraged CNNs trained on labeled RGB data using PyTorch, TensorFlow, and Keras to automate skeletal tracking and extract behavioral phenotypes. ### Full-Stack(AI/ML & Backend) Engineer II @ Jio Platforms Limited (JPL) Jan 2023 – Jan 2023 | Bengaluru, Karnataka, India • Improved model implementation and deployment timelines by 40% by automating preprocessing, feature engineering, training, and evaluation pipelines in Amazon SageMaker, with MLflow tracking. Improved liquidity planning by automating forecasting workflows and validating models via backtesting. • Achieved 3x faster load times on storefront product pages through full-stack data optimization, including PostgreSQL indexing and query refactoring, MongoDB schema restructuring with sharding and denormalization, and the implementation of a Redis-based distributed caching layer. • Constructed a resilient, Kafka-based messaging pipeline supporting event-driven workflows, real-time data ingestion, and state propagation across decoupled microservices through scalable producer-consumer patterns and fault-tolerant message queues. • Refactored and migrated a monolithic SAP Hybris platform to a microservices-driven, cloud-native Java Spring stack, enabling enhanced maintainability, reduced deployment friction, and improved operational agility. ### Associate Full-Stack(AI/ML & Backend) Engineer II @ HighRadius Jan 2021 – Jan 2023 | Hyderabad, Telangana, India • Increased dollar recovery by 12% and reduced manual research effort by 25%, building classification models (LightGBM, Random Forest) with 85% recall for fraud detection, ensuring SHAP interpretability. • Automated low-dollar deduction write-offs and prioritized research worklists through AI action suggestions, accelerating dispute resolution and improving operational efficiency. • Led a 10-engineer cross-functional team across Order-to-Cash products (backend + ML); boosted software quality by coaching developers, implementing code review standards, and optimizing Jenkins CI/CD workflows, achieving 40% fewer early-stage defects and 30% faster deployments. • Delivered Genesys Cloud VoIP integration for Thomson Reuters under SLA, reducing system failures by 15% and enabling on-time product launch. • Conducted real-time production monitoring using Elasticsearch to analyze multi-GB logs and identify deployment issues; monitored AWS infrastructure (EC2, S3, Auto Scaling Groups) and supported Snowflake ELT/ETL efforts by writing SQL queries for enterprise data analytics. ### Associate Software Engineer I @ HighRadius Jan 2019 – Jan 2021 | Hyderabad, Telangana, India • Collaborated with client finance teams, driving presentations to align model outputs with business workflows, ensuring adoption and trust in AI-driven fraud detection. • Optimized PostgreSQL performance by indexing high-traffic tables and restructuring expensive join operations using join reordering, filter pushdown and selective projections, reducing query execution time by 25%. • Reproduced, debugged, and fixed client-reported issues in staging environments by analyzing application logs and test conditions; performed QA and impact testing to ensure fix integrity and regression safety, resulting in a 45% improvement in overall system stability. • Automated B2B document aggregation and processing (POD, BOL, invoices, etc.) by leveraging Selenium, OCR processing, Captcha handling via Death By Captcha API, and rule-based automation pipelines using regex matching to extract critical data (delivery dates, payment terms, invoice details) from payment portals, emails, and SAP systems, enabling automated tracking for financial workflows. ## Education ### Master of Science - MS Brown University Jan 2023 – Jan 2025 ### Bachelor of Technology Kalinga Institute of Industrial Technology, Bhubaneswar Jan 2016 – Jan 2020 ### 11th & 12th D.A.V. Public School, Pokhariput ### 10th Board Buxi Jagabandhu English Medium School(B.J.E.M.S) ## Contact & Social - LinkedIn: https://linkedin.com/in/subham-kumar-das - GitHub: https://github.com/lost-particles - Email: mailto:subhamkumardas407@gmail.com --- Source: https://flows.cv/subhamkumardas JSON Resume: https://flows.cv/subhamkumardas/resume.json Last updated: 2026-04-18