Building AI systems from data → retrieval → reasoning → APIs → deployment.
I'm Madanala Abhishek Varma, a B.Tech CSE graduate focused on AI engineering, with hands-on experience across machine learning, data analytics, GenAI and agentic systems.
I started with frontend development, moved into machine learning and data pipelines, and gradually shifted toward building AI applications that combine retrieval, stateful workflows, APIs, evaluation and deployment.
I enjoy working across the full system rather than only the model layer.
| DATA | ML | RAG | AGENTS | APIs | DEPLOYMENT |
| Pipelines | Models | Retrieval | Workflows | Backend | Cloud |
|
Stateful AI workflows that coordinate retrieval, tools, reasoning and structured outputs.
|
Applications that combine video, speech, text and visual information.
|
|
End-to-end machine learning workflows from preprocessing to evaluation and inference.
|
Data pipelines and analytical applications built around Python, SQL and databases.
|
An agentic system designed to audit video content against retrieved advertising and compliance policies.
VIDEO
│
┌────────┴────────┐
▼ ▼
SPEECH ON-SCREEN TEXT
│ │
Whisper AWS Rekognition
│ │
└────────┬────────┘
▼
POLICY RETRIEVAL
FAISS + Embeddings
│
▼
LANGGRAPH WORKFLOW
│
▼
GROQ LLM
│
▼
PASS / FAIL + FINDINGS
│
▼
LANGSMITH TRACE
Stack
LangGraph RAG Whisper AWS Rekognition RAGAS
FAISS Sentence Transformers Groq FastAPI Docker LangSmith
Highlights
- Multimodal speech and on-screen text extraction
- Retrieval-grounded compliance reasoning
- Stateful LangGraph orchestration
- Structured findings and severity classification
- Automated testing and RAG evaluation
- LangSmith tracing and performance analysis
A financial intelligence platform covering 92 Nifty 100 companies, combining ETL, financial analytics, screening, peer benchmarking, ML-based analysis and reporting.
12 SOURCE DATASETS
│
▼
ETL + VALIDATION
│
▼
SQLITE DATA LAYER
│
┌────┼─────┬──────────┐
▼ ▼ ▼ ▼
KPI SCREEN PEERS ML/NLP
│ │ │ │
└────┴─────┴──────────┘
│
▼
FASTAPI + STREAMLIT
│
▼
FINANCIAL REPORTS
Stack
Python Pandas SQL SQLite FastAPI
financial-analysis Streamlit Scikit-learn KMeans
Highlights
- 12-source ETL pipeline with data-quality validation
- 30+ financial metrics
- 6 preset screeners
- 11 peer groups for benchmarking
- KMeans-based analytical clustering
- 19-endpoint FastAPI backend
- 8-screen Streamlit dashboard
- 172 automated tests
- Automated company, sector and portfolio reports
An agentic research system designed to search and reason over ArXiv research papers using Agentic RAG, with Exa Web Search as a fallback when local retrieval is insufficient.
USER QUERY
│
▼
QUERY ANALYSIS
│
▼
LOCAL ARXIV RETRIEVAL
│
▼
RETRIEVAL EVALUATION
│
┌───────────┴───────────┐
│ │
SUFFICIENT INSUFFICIENT
│ │
│ ▼
│ EXA WEB SEARCH
│ │
└───────────┬───────────┘
▼
CONTEXT ASSEMBLY
│
▼
LLM REASONING
│
▼
GROUNDED RESPONSE
Stack
LangGraph LangChain ChromaDB OpenAI Embeddings
Retrieval Evaluation Exa-api LLMs OpenAI API
Key Engineering Work
- Retrieves relevant research content from the local ArXiv knowledge base
- Evaluates retrieved context before generating an answer
- Uses Exa Web Search as a fallback when local retrieval is insufficient
- Combines retrieved evidence before LLM-based reasoning
- Uses an agentic workflow to decide when additional search is required
- Focuses on retrieval quality and grounded responses rather than relying only on the LLM
An end-to-end mutual fund analytics platform developed during my Data Analyst internship, processing 87K+ AMFI India records across 40 fund schemes via data ingestion, ETL, database modeling, exploratory analysis, performance analytics, risk analysis and Power BI reporting.
RAW FINANCIAL DATA
│
▼
DATA INGESTION
│
▼
CLEANING + TRANSFORMATION
│
▼
SQLITE STAR SCHEMA
│
┌────┼─────────┐
▼ ▼ ▼
EDA PERFORMANCE RISK
ANALYTICS ANALYTICS
│ │ │
└───────┴──────────┘
│
▼
POWER BI DASHBOARD
Stack
Python Pandas Numpy SQL SQLite
Data visualization plotly Power BI
Analytics
CAGR Sharpe Sortino Alpha Beta
Maximum Drawdown VaR CVaR HHI
Key Engineering Work
- Built ETL workflows for mutual fund datasets
- Cleaned and transformed NAV, transaction, AUM and SIP data
- Designed a SQLite star schema for analytical workloads
- Implemented performance and risk-adjusted financial metrics
- Developed analytical SQL queries for fund and investor analysis
- Built a risk-based fund recommender
- Created a 4-page interactive Power BI dashboard with drill-through and slicers
Repository · Power BI Dashboard
LangGraph LangChain CrewAI Autogen RAG FAISS
Sentence Transformers OpenAI Gemini Groq Hugging Face
Python Scikit-learn XGBoost Random Forest
TensorFlow NLP Classification Regression KMeans
Pandas NumPy SQL SQLite PostgreSQL
Matplotlib Seaborn ETL Power BI Streamlit
FastAPI REST APIs Docker Git GitHub
AWS Azure Streamlit React JavaScript Tailwind CSS
React
│
▼
Machine Learning
│
▼
NLP + Data Analytics
│
▼
Data Pipelines + SQL
│
▼
GenAI + RAG
│
▼
Agentic AI
│
▼
End-to-End AI Systems
The common thread:
Understand the problem → build the pipeline → connect the components → evaluate the system → deploy it.
|
|
| Role | Focus |
|---|---|
| Data Analyst Intern · Bluestock Fintech | Financial analytics · ETL · SQL · Risk metrics · Power BI |
| Frontend Developer Intern · Coreline Solutions | React · Tailwind CSS · AI API integration · Voice interfaces |
| AI/ML Intern · Edunet Foundation | NLP · TF-IDF · Random Forest · Gradient Boosting |
I'm interested in opportunities involving:
AI Engineering · Agentic AI · GenAI · Machine Learning · Data/ML Systems
DATA → RETRIEVAL → REASONING → SYSTEMS → DEPLOYMENT
Building things, breaking things, understanding why, and building them better.