Sayojya Patil

Sayojya Patil

Software Engineer · Dallas, TX

I build the backends for systems that watch real-world infrastructure: structural sensors, river flow, industrial batching plants. Python services that take in messy sensor data, run the analysis reliably, and give engineers results they can trust.

Open to remote, hybrid, or on-site roles across the US · willing to relocate
Experience & Research

Where I've worked

Jun 2025 – Present
Dearborn, MI

Research Software Engineer, Structural Health Monitoring

University of Michigan–Dearborn
Industry research partnership with HBK World
  • Built a FastAPI analysis service and React frontend that automate monitoring of structures from sensor data, implementing the lab's published research method and extending it to keep results consistent across runs.
  • Made onboarding configuration-driven, with a React setup wizard for configuring sensors and validating setups before each run.
  • Built a React, Three.js, and ECharts workspace with 3D views, analysis diagrams, trend graphs, and anomaly warnings.
  • Ran long analysis jobs in the background with asyncio and worker threads so the API stays responsive, and added data-quality checks before each run.
  • Used GitHub Copilot to scaffold the configuration modal and generate validation tests, then reviewed and extended them with pytest and Jest.
Jun 2025 – Present
Dearborn, MI

Research Software Engineer, Streamflow Prediction

University of Michigan–Dearborn
  • Cut experimentation overhead by 40% by refactoring research scripts into 20+ independent MLRun services on Kubernetes, packaged with Docker and released through GitHub Actions CI/CD.
  • Built a Python ingestion pipeline unifying four incompatible public APIs (USGS, PRISM, GridMET, GeoGLOWS) with async ingestion, schema validation, retry logic, and rate-limit handling, delivering 40+ years of daily data from 100+ river sites.
  • Built training pipelines for hybrid LSTM models that cut RMSE by 88% versus a physics-only baseline in evaluation.
  • Added KGE-based drift detection that scores model output against observed river flow, flags degradation, and triggers automatic retraining.
Jun 2023 – Jul 2024
Karnataka, India

Software Engineer

IDS Technologies
  • Cut report generation time by 30% for 100+ daily users with a batch reporting module that turns PLC production data into PDF reports.
  • Reduced complaint resolution time by 20% with a service-management workflow handling 500+ monthly cases end to end.
  • Shipped Kotlin Android features giving plant operators real-time batch status, material weights, alarms, and production metrics.
  • Diagnosed production issues across customer plants by tracing application logs, API responses, and database records.
Sep 2022 – May 2023
Karnataka, India

Backend Engineer Intern

Vayavya Labs
  • Built and Dockerized a Python/Flask REST backend with session-based auth (Flask-Login, Werkzeug, MySQL/SQLAlchemy) powering the DDGen driver-generation platform.
  • Reduced manual device-configuration effort by 50% through automated configuration and driver-generation workflows.
  • Profiled and optimized the multi-file processing pipeline, cutting backend execution time by 18%.
  • Added a CI check that ran the API test suite on every code change, backed by Postman tests and API documentation.
Projects

Things I've built

Pulse – Distributed Uptime Monitoring Platform

Uptime monitoring that scales out without double work. A FastAPI control plane queues HTTP health checks in Redis, and distributed locking makes sure each one runs exactly once: zero duplicates across 100K+ queued jobs. Scaling workers from 2 to 8 raised throughput 3.2x to 137 checks/sec at sub-100 ms P95. Deployed on AWS ECS/Fargate with RDS, ElastiCache, and CloudWatch.

  • FastAPI
  • PostgreSQL
  • Redis
  • Docker
  • AWS ECS/Fargate
  • CloudWatch

AI-Powered Loan Processing Platform

Authenticated Django REST backend for loan submissions, where document-text extraction and rule-based underwriting run as Celery jobs that start only after the database commit, so a retry can never process a loan twice. Owner-scoped access, duplicate-submission protection, non-root Docker, and 16 tests in GitHub Actions. Load tests before auth was added: 17,256 requests, zero failures, 297 requests/sec at 21 ms P95.

  • Python
  • Django REST Framework
  • PostgreSQL
  • Redis
  • Celery
  • Docker
  • GitHub Actions

Chicago Food Safety BI Platform

Public data, end to end. A Python ETL pages, validates, and upserts 50,000 City of Chicago inspection records into PostgreSQL, with retries and rollback so a failed page never leaves half-written data. FastAPI serves SQL-aggregated metrics to a React + Recharts dashboard: five KPIs, failure rates by risk level, and ZIP-code trends. 16 tests in GitHub Actions CI.

  • Python
  • FastAPI
  • PostgreSQL
  • React
  • Recharts
  • GitHub Actions

Clinical AI — Note Summarizer

Handwritten and scanned clinical records in, structured summaries out. A FastAPI pipeline uses GPT-4o Vision and validates every summary against PydanticAI schemas before saving it, behind JWT/OAuth2 auth, with PyTest coverage across extraction, auth, and storage.

  • FastAPI
  • PydanticAI
  • GPT-4o Vision
  • JWT/OAuth2
  • PostgreSQL

Warehouse Management System

Warehouse backend where stock can't go missing: capacity-limited bins, atomic transfers, and an append-only audit trail of every movement. Velocity-based (A/B/C) slotting, allocation that favors fast-moving zones, walk-order pick lists, and stale allocations rejected at shipping with HTTP 409.

  • Python
  • FastAPI
  • SQLAlchemy 2.0
  • PostgreSQL
  • Docker Compose
  • pytest

RAG Poisoning vs. Prompt Injection

Team security study of a RAG pipeline (MiniLM embeddings, FAISS top-5 retrieval, FLAN-T5) on 50 questions and a 500-passage corpus. Corpus poisoning cut answer accuracy from 88% to 58% (26% attack success) versus 4% for prompt injection. A keyword filter blocked the tested injections, while a retrieval-score anomaly detector made poisoning worse (26% to 34%).

  • Python
  • Sentence-Transformers
  • FAISS
  • flan-t5

Sequence Model Benchmark for Stock Forecasting

Team benchmark of seven sequence models (RNN, LSTM, GRU, Transformer, CNN-BiLSTM-Attention) against random-walk baselines on noisy stock data. Gated recurrent models beat the Transformer on short windows, reaching 78% directional accuracy against a 50% baseline.

  • Python
  • TensorFlow
  • Keras
  • pandas
  • NumPy

Flight Delay Analysis

PySpark pipeline over 10M+ flight records, from ingestion and cleaning to SQL aggregation, with visualizations of delay patterns by carrier, route, and time window.

  • Python
  • PySpark
  • SQL
  • pandas
Toolkit

What I work with

Languages
  • Python
  • TypeScript
  • JavaScript
  • SQL
  • Java
  • Go
  • Kotlin
Backend
  • FastAPI
  • Django
  • Flask
  • SQLAlchemy
  • asyncio
  • Celery
  • Redis
  • REST
  • JWT/OAuth2
Frontend
  • React
  • Next.js
  • Three.js
  • ECharts
  • Tailwind
  • Recharts
  • Vite
Data & AI
  • Anthropic API
  • OpenAI API
  • GPT-4o Vision
  • PydanticAI
  • PyTorch
  • TensorFlow
  • pandas
  • PySpark
Databases
  • PostgreSQL
  • ClickHouse
  • MySQL
  • MongoDB
  • InfluxDB
Cloud & DevOps
  • AWS (ECS/Fargate, RDS, ElastiCache, CloudWatch)
  • GCP
  • Docker
  • Kubernetes
  • MLRun
  • GitHub Actions
  • CI/CD
  • Linux
Testing
  • PyTest
  • Jest
  • Postman
AI-Assisted Development
  • Claude Code
  • Cursor
  • GitHub Copilot
Education

Where I studied

M.S., Computer and Information Science

University of Michigan–Dearborn
Aug 2024 – Apr 2026 · GPA 3.77 / 4.0

B.E., Computer Science and Engineering

Visvesvaraya Technological University, India
Jul 2019 – Jun 2023 · GPA 3.50 / 4.0
Contact

Looking for a software engineer? Let's talk.