poojithak/
Initialising experience
Home 01 About 02 Stack 03 Work 04 Journey 05 Achievements 06 Contact 07
Available · Open to AI Roles & Research

Building intelligent systems at the frontier of AI.

Elite AI Engineer · Full Stack Developer · Researcher · Systems Builder

Naga Poojitha Kanchukommala — building production-grade AI systems, Springer-published research, and shipping engineering excellence end-to-end. Specialised in LLMs, RAG, Agentic AI, and Deep Learning.

Claude API MCP LangChain PyTorch Next.js FastAPI Docker Springer 2025
scroll
About

Engineering the future
one system at a time.

I am an AI engineer and researcher operating at the intersection of language models, full-stack systems, and scientific rigour — turning frontier research into products that ship.

My approach is product-first and research-grounded. I build AI systems that are measurable, reliable, and deployed — not weekend prototypes. From conversational SQL agents powered by Claude & MCP to deep-learning pipelines for clinical diagnostics, every system I ship is benchmarked, documented, and built to scale.

My Springer-published research on combinatorial optimisation reflects how I approach engineering: rigorously, quantitatively, and always against strong baselines. I bridge AI/ML, full-stack development, and hardware engineering — a rare cross-domain fluency in a world that increasingly demands it.

01
AI Engineering
RAG pipelines, Agentic AI workflows, LLM orchestration, MCP integrations.
02
Full-Stack Systems
Next.js + FastAPI products with type-safe APIs and modern UX.
03
Deep Learning
PyTorch and TensorFlow models for medical imaging and predictive modeling.
04
Research
Springer-published — optimisation algorithms benchmarked against the field.
Tech Stack

An ecosystem
of modern engineering.

A curated stack covering AI/ML systems, full-stack web, deep learning, and developer infrastructure — selected for reliability and shipping speed.

🐍
Python
Language
C++
Systems
JavaScript
Language
TS
TypeScript
Type-safe
React
UI Library
Next.js
Framework
🎨
Tailwind CSS
Styling
🎬
Framer Motion
Animation
FastAPI
API
📊
Dash
Dashboards
📈
Plotly
Visualisation
🔥
PyTorch
DL Framework
🧠
TensorFlow
DL Framework
Claude API
LLM
MCP
Protocol
🤖
AI/ML Systems
Pipelines
🧬
Deep Learning
Modeling
💬
NLP
Language
🐳
Docker
Containers
🐘
PostgreSQL
Database
🗃
SQLite
Database
Git
Version Ctrl
📉
Data Viz
Insights
Featured Work

Selected projects
shipped with intention.

Project 01
AI · Claude · MCP

AI-Powered Natural Language to SQL Chatbot.

A conversational agent that turns plain-English questions into production-grade SQL queries — built on Claude API with Model Context Protocol (MCP) for live database connectivity. Enables non-technical teams to query operational data in natural language with safety guardrails and schema-aware responses.

~3s
Avg query time
95%
Query accuracy
MCP
Protocol-native
Claude APIMCP PythonFastAPI PostgreSQLSQLite
nl-sql · claude · mcp
▸ user query
Show me top 5 customers by revenue last quarter
→ schema resolved · 3 tables · 12 joins inferred
SELECT c.name, SUM(o.amount) AS revenue
FROM customers c JOIN orders o
  ON c.id = o.customer_id
WHERE o.created_at >= '2025-Q4'
GROUP BY c.name
ORDER BY revenue DESC LIMIT 5;
Diagnosis
Venous Ulcer
Stage III · Healing
Confidence
92.4%
Model
ResNet-50
Fine-tuned
Classes
Pressure87%
Venous92%
Diabetic89%
Project 02
Deep Learning · Healthcare

Chronic Wound Classification using Deep Learning.

A CNN-based clinical decision-support tool that classifies chronic wound types from images — pressure ulcers, venous ulcers, diabetic ulcers, and surgical wounds. Built with transfer learning on ResNet-50, augmented with domain-specific preprocessing for variable lighting and occlusion in clinical photography.

92%+
Classification acc.
4 cls
Wound types
CNN
Transfer learning
PyTorchTensorFlow ResNet-50OpenCV Python
Project 03
Data Visualisation · Public Health

COVID-19 Interactive Visualisation Dashboard.

A real-time interactive dashboard for tracking and analysing COVID-19 data across geographies and time. Built with Plotly Dash, featuring choropleth maps, time-series forecasting plots, and demographic breakdowns — designed for both researchers and policy makers to surface signal from noisy public-health data.

200+
Countries tracked
Live
Real-time updates
8+
Chart types
DashPlotly PandasPython REST APIs
Cases
762M
Recovered
734M
Active
21.2M
— Global daily cases · 14d trend
Help me plan my literature review on RAG architectures this week.
research assistant
I've drafted a 4-day plan:
  • Day 1 — Survey 6 foundational papers (auto-collected)
  • Day 2 — Compare retrieval strategies
  • Day 3 — Draft synthesis section
  • Day 4 — Citations + bibliography (.bib generated)
Add the new Anthropic paper.
indexing
arXiv:2503.xxxxx queued · added to Day 1 reading.
Project 04
AI Assistant · Productivity

AI Assistant for Academic & Research Task Management.

An LLM-powered research assistant that helps researchers plan reading lists, summarise papers, track deadlines, and generate citation bibliographies. Built with LangChain orchestration, Next.js frontend, and FastAPI backend — designed as the AI co-pilot researchers wished they had during their PhD.

10×
Faster lit review
Auto
.bib generation
Multi
LLM orchestration
LangChainLLMs Next.jsTypeScript FastAPIPostgreSQL
Project 05
Mobile · Social Impact

Blood Donation Android Application.

A native Android app connecting blood donors with recipients in real-time. Geolocation-based matching, blood-group filtering, push notifications, and verified donor profiles — built to reduce the average donor-recipient connection time in critical situations.

Real-time
Donor matching
8 grps
Blood types
GPS
Geo-matched
AndroidJava / Kotlin FirebaseMaps API
LifeDrop
Donors nearby23
Blood groupO+
Distance2.4 km
Avg response12 min
Request Donor →
Journey

A chronology
of engineering & research.

Mar 2026 · Recent
IBM RAG & Agentic AI · Professional Certificate
IBM Skills Network · Coursera · 8-course series
Completed IBM's professional certification covering Retrieval-Augmented Generation, agentic workflows, vector databases, and production LLM deployment patterns.
2025 · Publication
Springer Nature · CCIS Vol. 2738
Communications in Computer and Information Science · Peer-reviewed
Published "Optimizing Charger Placement for Q-Coverage Using Blackhole Algorithm" — demonstrated 12–18% reduction in charger count vs. GA and PSO baselines on WSN topologies up to 500 nodes.
SpringerScopus-IndexedOptimisation
2024 — 2025 · Building
AI Engineering Projects
Self-directed · Production deployments
Built end-to-end AI systems including the Claude+MCP NL→SQL chatbot, chronic wound classification deep learning pipeline, and AI research assistant — all instrumented with metrics and shipped as deployable services.
RAGMCPDeep Learning
B.Tech. Electronics & Communication Engineering
Amrita Vishwa Vidyapeetham, Coimbatore
Pursuing B.Tech in ECE with concentration in VLSI Design, Digital Signal Processing, and Embedded Systems. Maintained academic excellence while building production AI systems and publishing research.
VLSIDSPEmbedded
By the Numbers

Measurable impact,
built into every system.

0%+
Hallucination Reduction
0%
DL Classification Acc.
0%
WSN Charger Reduction
0
Featured Projects
0
Springer Publication
0
IBM Certifications
✦ Featured Publication · Springer Nature

Optimizing Charger Placement for Q-Coverage Using Blackhole Algorithm

Communications in Computer and Information Science · Vol. 2738 · 2025 · Scopus-Indexed

Available for Opportunities

Let's build
something remarkable.

Open to graduate research, full-time AI engineering roles, freelance, startup collaboration, and conversations on cutting-edge LLM & Agentic AI work. Responses typically within 24 hours.

Direct email
hello@poojithak.com
— Send a Message
Encrypted & private
Hit send — I'll reply to your email.