Open to software, AI/ML and data roles

G Pradeep Kumar

Software Engineer, Backend & AI

I build backend services and AI applications end to end — LLM agents, fintech microservices and computer-vision models, all tested and shipped through CI.

See my projects Download resume
Hey there! 👋
  • 7projects on GitHub
  • 3roles
  • 4certifications
  • 2025B.Tech CSE

👋 Hover over me to say hi, or tap for more.

Watch a request move through my code.

Each of my four newest projects, step by step — from the request coming in to the result going out. Pick a project to replay it.

AI Support Agent
How a guest message becomes a grounded reply
Running

    Code that ships, backed by data.

    I'm a Computer Science graduate from Dayananda Sagar University who builds backend services and AI applications from the first schema to the CI pipeline. Every project here is tested, containerised and running in GitHub Actions.

    Before that, I wrote Python automation at an EdTech startup, then moved into leadership and operations — leading a team that won a 5-year institutional partnership, and running CRM analytics with SQL. That mix taught me to tie engineering work to a metric someone cares about.

    I'm looking for Software Development, AI/ML or Data roles where I can own a problem end to end.

    Portrait of G Pradeep Kumar
    G Pradeep Kumar · Bengaluru
    Based in
    Bengaluru, India
    Degree
    B.Tech CSE, 2025
    Languages I code in
    Python, Java, TypeScript, SQL
    Focus
    Backend, GenAI, ML
    Speaks
    English, Kannada, Telugu, Hindi

    Things I've built.

    All seven repositories on my GitHub. Open any project for the details.

    • AI Support Agent

      New

      An LLM customer-support agent for a hotel front desk. It answers from a knowledge base with RAG and calls tools to look up bookings, open tickets and hand off to a human.

      PythonFastAPIOpenAI & Anthropic APIsRAGPostgreSQLDockerKubernetes
      18/18
      eval scenarios passing in CI
      Code
      • Provider-agnostic LLM layer for OpenAI, Anthropic and an offline mock, with retries, versioned prompts and guardrails.
      • Agent loop with a step budget and tool-error recovery; running out of steps escalates to a human instead of looping.
      • HMAC-signed, idempotent inbound webhook for WhatsApp/SMS providers, with replay protection.
      • LLM extraction of guest emails into schema-validated booking JSON, plus a CSV/Excel booking importer.
      • 18-scenario eval suite and 24 pytest tests gate every merge in GitHub Actions.

    • KYC Orchestrator

      New

      A fintech verification microservice. One API call checks identity, sanctions and address vendors in parallel, scores the risk, and approves, rejects or sends the case to manual review.

      TypeScriptNode.jsExpressPostgreSQLJestDockerKubernetes
      29
      Jest tests, incl. PostgreSQL in CI
      Code
      • Parallel vendor fan-out keeps latency at the slowest vendor rather than the sum of all three.
      • Timeouts, exponential-backoff retries with jitter, and a TTL/LRU cache that cuts repeat vendor calls.
      • Never auto-approves on missing data: a vendor timeout forces manual review.
      • Multi-tenant API keys and idempotency keys make client retries safe.
      • Transactional outbox delivers HMAC-signed webhooks with retries and dead-lettering.

    • Defect Vision

      New

      An industrial visual-inspection pipeline that classifies surface defects, flags unseen anomalies and draws where the defect is — then shrinks the model for edge devices.

      PythonPyTorchOpenCVONNXFastAPI
      99.75%
      validation accuracy, 0.996 macro-F1
      Code
      • CNN trained on a 4,000-image, imbalanced synthetic dataset of scratches, spots and cracks.
      • Autoencoder trained only on good parts catches 90.4% of defects at a 1% false-alarm rate.
      • Localises defects with residual thresholding, morphology and contours into boxes and masks.
      • int8 quantization shrinks the model from 354 KB to 123 KB at equal accuracy; ONNX export with a parity test.
      • FastAPI inspection endpoint returns a pass/reject verdict per part image.

    • Shipment Tracking Service

      New

      A logistics REST API for booking shipments and tracking them through their delivery lifecycle, built with a layered controller–service–repository design.

      Java 21Spring Boot 3Spring Data JPAPostgreSQLFlywayDocker
      8-state
      delivery lifecycle as a state machine
      Code
      • Illegal status jumps are rejected with 409; failed deliveries can retry, return or cancel.
      • Estimated delivery dates from Indian pincode regions, skipping Sundays.
      • Optimistic locking so concurrent scanner updates can't overwrite each other.
      • Full tracking history in one query with @EntityGraph — no N+1; paginated, filterable listing.
      • Flyway migrations, RFC 7807 errors, Actuator health probes, JUnit 5 and MockMvc tests.

    • Personal Finance Tracker API

      A REST API for tracking income, expenses and budgets, with authentication and per-user data isolation.

      Node.jsExpressMongoDBJWT
      7
      REST endpoints across 2 models
      Code
      • JWT authentication with bcrypt password hashing.
      • 8 income/expense categories with full CRUD and filtering.
      • Ownership-scoped queries keep each user's data isolated.
      • Server-side MongoDB aggregation for income, expense and budget analytics.

    • Credit Risk Prediction Model

      A machine-learning model that predicts loan default risk from applicant data, evaluated for class imbalance.

      PythonScikit-learnPandas
      85.3%
      ROC-AUC on 5,000 applicants
      Code
      • Compared Logistic Regression and Random Forest on 5,000 applicant records.
      • 80% accuracy, with precision, recall and F1 to account for class imbalance.
      • Top predictors: credit score, debt-to-income and income.

    • Transactions Analytics Dashboard

      An interactive dashboard for exploring payment transactions by customer, category and channel.

      PythonPandasStreamlit
      6,000+
      transactions across 799 customers
      Code
      • Covers 8 spending categories and 4 payment channels with dynamic filters.
      • 4 live KPIs: volume, transaction count, success rate and average transaction size.

    Where I've worked.

    Engineering, team leadership and operations — newest first.

    1. Oct 2025 – Feb 20265 months

      Assistant Manager – Operations

      Yavhi (Aarde Projects)
      • Managed CRM operations end to end, using data analysis to track leads, customer engagement and pipeline health.
      • Wrote SQL queries and used Excel to produce insights that improved conversion rates and informed operational decisions.
      • Built dashboards and kept data accurate, supporting faster, evidence-based decisions across the team.
    2. Feb 2025 – Sep 20258 months

      Team Lead

      LearnFlu, early-stage startup
      • Owned a team driving product adoption for an AI/ML learning program, breaking engagement and growth problems into cross-functional tasks.
      • Used performance data to prioritise initiatives — recognised as Top ATL (May) and Top Team Lead (July).
      • Led the customer negotiations and execution that secured a 5-year institutional partnership.
    3. Sep 2023 – Dec 20234 months

      Python Developer Intern

      BlendVidya EdTech
      • Built backend features and automation scripts in Python and MySQL to streamline data-processing workflows.
      • Designed and shipped an end-to-end train ticket booking automation system that cut manual effort.
      • Debugged issues and optimised database queries for better performance and reliability.
      • Worked in a cross-functional team using Git in a fast, iterative environment.

    What I work with.

    Languages

    PythonJavaTypeScriptJavaScriptSQL

    Frameworks

    FastAPISpring BootNode.js (Express)React

    Databases

    PostgreSQLMySQLMongoDB

    AI & machine learning

    OpenAI & Anthropic APIsRAGAI agents & tool callingLLM evaluationPyTorchOpenCVONNXScikit-learnPandas

    Tools & cloud

    GitDockerKubernetesGitHub ActionsAWS

    Certifications

    • Tata GenAI Powered Data Analytics Job SimulationForage · Verification 6ac7eb64df29216275787cc9
    • Full Stack DeveloperOne Roadmap Skill Certification · CERT-7C2DC4F2
    • React.JS DeveloperOne Roadmap Skill Certification · CERT-4AC85013
    • AI and Data ScientistOne Roadmap Skill Certification · CERT-3656FDA3

    Education

    • B.Tech, Computer Science and EngineeringDayananda Sagar University, Bengaluru
    • Pre-UniversityDeeksha PU College

    Let's build something.

    I'm open to Software Development, AI/ML and Data roles. Email me or grab the full resume.

    Email me Download resume
    +91 97312 99940 LinkedIn GitHub