manikanta.
AI & Full Stack Engineer

The demo is the easy part.

I build AI and full stack systems, then measure them honestly. An agent that writes SQL and repairs its own mistakes. An assistant that refuses to guess. A billing system that cannot oversell stock. All three live, all three tested.

Selected workGet in touch
GitHubLinkedIn
Manikanta Pudi
B.E. CSE (AI & ML) · 2025Open to roles
0
Unsafe queries reached the database
0.92
Threshold accuracy, up from 0.83
0.97+
R², suitability model
6.7M+
Records analysed
01 — Selected work

Systems I built, and what they measure

Agentic AI · 2026AI / ML

SQL Analyst Agent

Ask a question in English, get the answer and the SQL that produced it. A LangGraph agent writes a query, runs it, reads the structured error, and repairs itself using a strategy chosen per SQLSTATE. It stops after three attempts and cannot do damage, because the credential it holds cannot write.

  • LangGraph
  • MCP
  • FastAPI
  • PostgreSQL 17
  • sqlglot
  • Groq
  • Langfuse
Read the case study →GitHubLive app
118
Tests
0
Unsafe SQL reaching the database
12
Error types, each with a strategy
3
Credentials, least privilege
Full stack · 2026Full stack

Anvil Hardware

A point of sale and stock system for a hardware store. Django and DRF over PostgreSQL, React on the counter. Built around one rule: stock can never be oversold, even when two cashiers bill the last unit at the same instant.

  • Django 5.1
  • DRF
  • PostgreSQL 17
  • React 19
  • Vite 8
  • JWT
  • pytest
Read the case study →GitHubLive app
7
Tables, PROTECT on FKs
3
Django apps
2
Roles, enforced server side
400ms
Search debounce at the counter
Hybrid RAG · Aug 2026AI / ML

Scholarship Assistant

Answers eligibility questions over 44 Indian government scholarship schemes. Eligibility is numbers, and embeddings cannot compare numbers, so the rules are filtered in SQL before retrieval runs at all. Every answer is either grounded in a source or refused.

  • FastAPI
  • PostgreSQL
  • ChromaDB
  • bge-small-en-v1.5
  • Groq
  • RAGAS
  • React 19
Read the case study →GitHubLive app
0.92
Threshold accuracy, up from 0.83
0
Invented figures, down from 7.5%
81%
Ineligible schemes naive RAG surfaced
48
Tests
ML + Backend · Feb 2026AI / ML

EcoPackAI

Ranks 25 packaging materials across 13 product categories by suitability, cost and carbon impact. A composite eco-score makes the ranking readable; eleven REST endpoints and a normalised PostgreSQL schema make it usable.

  • Flask
  • XGBoost
  • scikit-learn
  • PostgreSQL
  • SQLAlchemy
  • Render
Read the case study →
0.98+
R², CO₂ impact
2,275
Training samples
15
Engineered features
11
REST endpoints
Front end · In progressFront end

MediCare HMS

The patient-facing side of a hospital system: browse doctors, book appointments, register, and a protected dashboard. Refactored out of a duplicated HTML/Bootstrap site into a React component architecture with custom hooks for fetching and debounced search. Backend is next.

  • React 19
  • Vite 8
  • React Router 7
  • Bootstrap 5.3
  • Axios
  • Vercel
Read the case study →GitHubLive app
10
Routes, 5 protected
8
Breakpoints, no h-scroll
2
Custom hooks
500ms
Search debounce
02 — Experience

Where I've worked

Dec 2025 — Feb 2026
Remote, India

Data Science Intern

Infosys Springboard
  • Built an AI packaging recommendation engine across 25 materials and 13 categories, returning ranked suitability, cost and CO₂ predictions in real time.
  • Trained Random Forest and XGBoost on 2,275 samples with 15 engineered features, reaching 0.97+ R² suitability and 0.98+ R² CO₂ estimation.
  • Served the models through a Flask REST API with joblib model loading, deployed on Render with managed PostgreSQL.
Jan 2025 — Jun 2025
Hyderabad, India

Data Analyst Intern

Techmatrics Solution
  • Analysed 6.7M+ retail transactions and showed that 20% discounting cut net revenue despite higher volume, and delivered that as a pricing recommendation.
  • Wrote multi-table JOINs, CTEs and window functions against PostgreSQL, cutting query time and removing manual prep from recurring reports.
  • Built Tableau dashboards on revenue by category, customer segment and region, plus reusable Python cleaning and validation functions.
03 — Stack

What I work with

Languages & core

Python · SQL · JavaScript · Data Structures & Algorithms · OOP · REST API design · System design basics · Statistics & probability

AI & LLM

LangChain · LangGraph · MCP · Google Gemini · Groq · ChromaDB · Langfuse · RAGAS · Hybrid retrieval · Cross-encoder reranking · Retrieval evaluation

ML & data

PyTorch · scikit-learn · XGBoost · Pandas · NumPy · Feature engineering · Tableau

Backend

Django · Django REST Framework · FastAPI · Flask · JWT auth · Pydantic · Uvicorn · Middleware & rate limiting · Transactions & row locking · Pytest

Front end

HTML5 · CSS3 · React · Vite · Next.js · TypeScript · Tailwind CSS · Bootstrap · Responsive layouts · Streamlit

Databases

PostgreSQL · MySQL · SQLAlchemy ORM · Schema design · Query optimisation · ChromaDB

DevOps & cloud

Docker · Docker Compose · Git · GitHub Actions · CI/CD · Google Cloud Run · Render · Vercel · Linux

04 — About

I care about the number after the demo.

A retrieval system that answers well in a notebook and drifts in production isn't finished. So I build the evaluation harness first: a 22-question dataset, LLM-as-judge scoring for faithfulness and citation accuracy, and quality gates in CI that refuse the deploy when the metric degrades.

Six months of paid analyst work taught me the other half: the model is only as useful as the question it answers. Before that, a YOLOv8 terrain-identification paper and a traffic-sign detector at 92% accuracy. Now: Hyderabad, and open to AI and full stack engineering roles.

Research & certifications

  • Off-Road Terrain Identification using YOLOv8
    Journal of Electrical Systems — publication
    Dec 2023
  • NextLeap Data Analyst Fellowship
    Top Fellow recognition
    Jul 2025
  • SQL (Intermediate)
    HackerRank certification
    Mar 2025

Education

B.E. Computer Science — AI & Machine Learning
Chandigarh University, Mohali · CGPA 7.98/10 · 2021–2025

Final year project: real-time traffic sign detection with YOLOv8, 92% classification accuracy.

05 — Contact

Hiring for AI or full stack? Let's talk.

Résumé — downloadAI / ML EngineerRésumé — downloadSoftware Engineer