Founding Machine Learning & Backend Engineer at VelocitiPM LLC (2025-06-01 - Present)

As the Founding ML & Backend Engineer at VelocitiPM, I architected and implemented the core multi-agent AI engine responsible for 4x throughput improvement. This complex system utilized LangChain, LangGraph, and CrewAI for sophisticated agent orchestration, deployed using a high-performance FastAPI service. I recently completed the "Protocol Battle Arena," a deep-dive into network systems engineering. I built a benchmarking suite that pits REST, GraphQL, gRPC, WebSockets, and SSE against each other. Technically, I engineered a server orchestration system in Python that monitors CPU and Memory usage at the process level using 'psutil'. I also optimized a SQLite backend to handle over 100,000 activity logs using batch-insertion scripts, ensuring the database wasn't the bottleneck during high-concurrency throughput tests. This project demonstrates my ability to design high-performance microservices and quantify architectural trade-offs using real-world telemetry. I am a core contributor to Zerobrew, a Rust-based systems tool. My primary contribution involved refactoring the dependency resolution logic to support parallel batch installations. This work touched three separate crates within the workspace and required maintaining zero clippy warnings in a strict CI environment. My commitment to code quality is reflected in my high activity, with over 600 contributions logged in early 2026 alone. I specialize in Rust systems engineering, parallel I/O, and maintaining atomic git hygiene in professional open-source projects. I recently built HimmiRouter, a high-performance LLM Inference Gateway. It uses a Python-based microservice architecture (FastAPI) and LangGraph to manage complex routing states. A key technical achievement was the implementation of a billing engine that uses PostgreSQL 'FOR UPDATE' locks to ensure atomic credit consumption, even during high-concurrency streaming. The system also includes a Semantic Cache using RedisVL to reduce latency and costs for repetitive prompts. The frontend is a sophisticated React dashboard with an "Obsidian" aesthetic, featuring real-time usage analytics via Recharts and a streaming AI Playground with a "Shadow Mode" feature for comparing model outputs side-by-side. For system-level operations, I standardized asynchronous orchestration by integrating C++ and Go micro-services, ensuring low-latency and high-reliability operations. A major achievement was co-leading the development of event-driven data pipelines using industry-standard tools like Kafka, AWS SQS, and SNS, coupled with AWS Lambda. This design drastically reduced our P95 latency by 85% for a 2,500-user pilot program, which was validated through rigorous A/B testing and async routing strategies. Furthermore, I established end-to-end MLOps practices, including CI/CD pipelines, a comprehensive model registry, and automated RAG re-indexing strategies. This operational discipline resulted in an 8x increase in our release cadence, directly contributing to a 400% boost in Daily Active Users (DAU), and significantly improving system resilience, cutting our critical incident resolution time from two hours to approximately 10 minutes.

Skills: Python, Go, C++, LangGraph, CrewAI, FastAPI, Kafka, AWS Lambda, AWS Kinesis, AWS SQS, Redis, MLOps, System Design, Observability. Core Accomplishments: Architected a **30-agent AI orchestration engine** (LangGraph, CrewAI) on AWS AppRunner, automating 80% of PM workflows and increasing engineering throughput **4×**.; Engineered an event-driven, async pipeline using **AWS Kinesis, SNS, and Lambda**, reducing P95 latency by **85%** and ensuring **99.98%** message delivery reliability.; Built a production-grade **MLOps platform** with SageMaker Pipelines and MLflow, automating retraining gates and boosting release cadence **8×**.; Implemented a robust **observability suite** (CloudWatch, X-Ray) that cut critical incident resolution time from 2 hours to **~10 minutes** via distributed tracing.; Designed a **Redis-based memory layer** for agentic state persistence, slashing inter-agent latency by **60%** and supporting **2.5×** higher concurrent user throughput.

Founding AI Engineer at TimelyHero, Dimes Inc. (2024-08-01 - 2025-06-01)

As the Founding AI Engineer at TimelyHero, Dimes Inc., I led the strategic migration of the user-matching service into a robust Java/Flask/gRPC microservice on Amazon EKS. The new architecture is built around Kafka for messaging and supports 100,000 concurrent WebSocket sessions at a P95 latency of 250ms, resulting in a 3.5x improvement in system stability. I designed and built Airflow-based RAG pipelines utilizing Pinecone, MongoDB, and S3, which dramatically cut data staleness from 48 hours to 30 minutes, improved semantic-match accuracy by 27%, and was crucial in securing $250K in enterprise contracts. I standardized IaC and CI/CD using Terraform and AWS CodePipeline/CodeBuild, which reduced deployment downtime by 35%, quadrupled release velocity, and doubled developer productivity. I also deployed a comprehensive monitoring stack (CloudWatch, Prometheus, Grafana), improving incident detection time by 70%. I implemented data back-pressure handling in the SNS/SQS + Kafka bridge, stabilizing ingestion throughput at 5,000 messages/s and preventing queue overloads under peak loads greater than 1 million events/hour.

Skills: Java, Flask, gRPC, Kafka, Airflow, RAG, Pinecone, MongoDB, Terraform, Kubernetes (AKS), Azure, System Design. Core Accomplishments: Spearheaded the migration of a legacy monolith to a **Java/Flask/gRPC microservice architecture** on AKS, scaling to support **100,000+** concurrent WebSocket sessions.; Designed and deployed **Airflow-orchestrated RAG pipelines** (Pinecone, OpenAI, MongoDB), reducing data staleness from 48 hours to **30 minutes** and driving **$250K+** in new enterprise contracts.; Standardized **Infrastructure-as-Code (IaC)** using Terraform and Kubernetes, reducing deployment downtime by **35%** and quadrupling the team's release velocity.; Optimized high-throughput ingestion pipelines with **Kafka back-pressure handling**, stabilizing throughput at **5,000 msgs/sec** under peak loads of 1M+ events/hour.; Implemented a comprehensive **monitoring stack** (Prometheus, Grafana) that improved proactive incident detection by **70%**, ensuring high availability.

Machine Learning Intern at Endimension Inc. (2024-04-01 - 2024-08-01)

As a Machine Learning Intern at Endimension Inc., I was responsible for training and optimizing computer vision models (TensorFlow, Keras) on Amazon SageMaker GPU clusters, achieving a 13% mAP and 17% IoU improvement across 3 million+ scans (8 TB dataset). I optimized inference by applying post-training quantization and mixed-precision techniques (ONNX Runtime + TensorRT) on SageMaker Endpoints, reducing P95 latency by 40% with negligible AUC drift (<0.5%). To optimize costs, I leveraged SageMaker Spot Instances and data-parallel training (Horovod + NCCL) on multi-node clusters, which cut GPU cost by 25% and accelerated the training iteration cycle by 1.8x. I built a serverless inference pipeline using Kinesis, Lambda, and SageMaker Endpoints, enabling 24x7 fault-tolerant processing and reducing inference P95 latency by 40%. I also implemented data preprocessing pipelines (AWS Glue + S3) to handle 8 TB of raw DICOM data, achieving 2.3x faster ingestion. I configured a monitoring stack (CloudWatch, Prometheus, MLflow tracking) for model metrics, which cut debug time by 60%.

Skills: TensorFlow, Keras, Computer Vision, Model Optimization, Quantization, ONNX Runtime, AWS SageMaker, CUDA, AWS Kinesis, AWS Glue, MLflow. Core Accomplishments: Trained and optimized large-scale vision models (TensorFlow, Keras) on **8TB of medical imaging data**, improving diagnostic mAP by **13%** and IoU by **17%**.; Deployed **ONNX-quantized inference endpoints** on SageMaker, reducing P95 latency by **40%** with negligible AUC drift (<0.5%) for real-time diagnostics.; Leveraged **SageMaker Spot Instances** and distributed training strategies to reduce GPU training costs by **25%** while accelerating iteration cycles by **1.8×**.; Architected a **serverless inference pipeline** using AWS Kinesis and Lambda, ensuring **24/7 fault tolerance** and scalable processing for high-volume image streams.; Configured a robust **model monitoring stack** (MLflow, Prometheus) to track drift and performance, reducing debugging time for production models by **60%**.

Graduate Research Assistant at JLiang Lab, Arizona State University (2023-09-01 - 2024-05-01)

As a Graduate Research Assistant at the JLiang Lab, Arizona State University, I developed and trained multi-modal Computer Vision (CV) models, including state-of-the-art architectures like Swin Transformer, DINOv2, I-JEPA, and CLIP. This was executed on Amazon SageMaker GPU clusters and resulted in a 24% increase in rare-disease recall and a 17% increase in overall F1 score on multi-institutional medical datasets. I achieved significant infrastructure efficiency by implementing Fully Sharded Data Parallel (FSDP) training on Amazon EKS with 8x A100 nodes, which cut training time 3.2x and reduced cloud cost by 40%. I configured distributed data pipelines via AWS Glue and S3, enabling the ingestion of 8 TB+ multimodal datasets and improving preprocessing throughput by 2.6x. I also built a multi-node GPU job scheduler with Slurm on EC2 Batch, automating node provisioning and improving GPU utilization by 35% across shared research workloads. I designed experiment tracking workflows using SageMaker Experiments and MLflow to ensure 100% reproducibility and reduce hyperparameter tuning time by 45%.

Skills: PyTorch, JAX, Computer Vision, Distributed Training (FSDP), Slurm, AWS EKS, AWS Glue, AWS S3, SageMaker, MLflow. Core Accomplishments: Developed state-of-the-art **multi-modal CV models** (Swin Transformer, DINOv2) on SageMaker GPU clusters, achieving a **24% increase** in rare-disease recall.; Implemented **Fully Sharded Data Parallel (FSDP)** training on Amazon EKS (8× A100s), reducing model training time by **3.2×** and cutting cloud compute costs by **40%**.; Engineered a **multi-node GPU job scheduler** using Slurm and EC2 Batch, automating provisioning and boosting cluster utilization by **35%** across shared workloads.; Built distributed data pipelines with **AWS Glue and S3** to ingest and preprocess **8TB+** of multimodal medical datasets, improving data throughput by **2.6×**.; Designed reproducible experiment tracking workflows with **SageMaker Experiments and MLflow**, reducing hyperparameter tuning time by **45%** and ensuring 100% reproducibility.

Machine Learning Researcher at SRM Advanced Electronics Laboratory (2021-12-01 - 2023-07-31)

In my role as a Machine Learning Researcher at SRM Advanced Electronics Laboratory, I developed a distributed kernel regression pipeline on Apache Spark (using Java and MLlib) for the prediction of non-invasive blood glucose from photoacoustic spectroscopy (PAS) signals. The pipeline achieved a highly accurate clinical-grade Mean Absolute Relative Difference (MARD) of 8.86% and RMSE of 10.94, with over 95% of predictions falling within the safe and accurate Clarke Error Grid Zones A and B, a testament to its reliability under high-noise, real-world sensor conditions. I engineered a real-time IoT data streaming pipeline leveraging AWS Greengrass for edge processing and AWS IoT Core for cloud messaging via MQTT, achieving an ultra-low-latency ingestion of less than 200ms for over 2,000 sensor readings per hour. I also designed robust ETL and data-quality workflows using Spark SQL, implemented automated validation metrics, and created evaluation frameworks to ensure production-grade signal integrity across various edge devices. This foundational work culminated in the co-authoring and publication of the research in **Scientific Reports** (Nature Portfolio), a Q1-ranked journal.

Skills: Apache Spark, SQL, Java, AWS Greengrass, AWS IoT Core, AWS MQTT, Regression Modeling, ETL, Research. Core Accomplishments: Developed a distributed **kernel regression pipeline** on Apache Spark (Java + MLlib), achieving a clinical-grade **MARD of 8.86%** for non-invasive glucose monitoring.; Engineered a real-time IoT streaming system using **AWS Greengrass and IoT Core**, enabling ultra-low latency ingestion (**<200ms**) for 2,000+ daily sensor readings.; Designed production-grade **ETL and data quality workflows** with Spark SQL, ensuring **95%+ signal integrity** across distributed edge devices.; Optimized cloud-to-edge messaging protocols via **MQTT**, ensuring reliable data transmission and reducing packet loss by **15%** in unstable network environments.; Co-authored and published this novel research in **Scientific Reports (Nature Portfolio)**, a Q1 journal, validating the system's clinical accuracy and architectural robustness.

Master of Science in Information Technology (AI/ML) from Arizona State University (2023-08-01 - 2025-05-01)

Location: Tempe, AZ, USA. Coursework: Digital Image Processing, Foundations of Statistical Machine Learning, Fundamentals of Machine Learning, Operationalizing Deep Learning, Image Analytics and Informatics, Advanced Operating Systems, Social Media Mining, Knowledge Representation and Reasoning, Cloud Computing, Statistical Machine Learning, Data-Intensive Distributed Systems for Machine Learning. Activities: SoDA: Software Developers Association, ACM Student Chapter, Linux Users Group, The AI Society at ASU, Hindu YUVA.

Bachelor of Technology in Computer Science from SRM University (2019-06-01 - 2023-06-01)

Location: Amaravathi, AP, India. Coursework: Biology, Chemistry, Calculus I, Calculus II, Basic Electronics, Digital Electronics, DSA in C, Physics, Statistics, Discrete Mathematics, OOPs in Java, Linear Algebra, Database Management Systems, Full Stack & Web Technologies, Formal Languages & Automata Theory, Economics, Computer Organization and Architecture, Introduction to Quantum Computations, Differential Equations, Operating Systems, Compiler Design, Data Warehousing and Data Mining, Computer Networks, Data Science, Software Engineering, Fundamentals of Neuro Linguistics Programming, Supply Chain Management, Managing Innovation and Startups, Cloud Computing, Big Data Analytics, Machine Learning. Activities: Founder @ Inventors Village, Founder @ Research Clan, Board Member @SRM Student Council, Board Member @ SRM Entrepreneurship Cell, Member @ GDSC.