CCA-F
Alin Verma
AI engineering, built to run in production.
I ship agentic AI — autonomous agents and RAG — grounded in the data engineering pipelines behind them. Most AI engineers can't build the data foundation; most data engineers can't ship the agents. I do both.
- Currently
- Data Engineer at Accelerize 360
- Stack
- Agentic Workflows · RAG · Data Engineering
- Elsewhere
- LinkedIn ↗GitHub ↗
Things I've built recently.
Each project below lives at its own subdomain. Click a card to open it.
Verifiable and current.
Four certifications, all in date. Click through to verify on the issuer's site.
Where I've put hours in.
Three roles, oldest at the bottom. Each line is a specific thing I owned, not a job description.
- Mar 2024—PresentFull-time
Data Engineer
Accelerize 360·Hyderabad
- Shipped an autonomous AI agent on the Claude API (Anthropic) that plans, drafts, and refines presentation decks through multi-turn chat. Deck creation dropped from 2+ hours to ~30 minutes.
- Built a text-to-SQL tool on Snowflake Cortex so marketers segment customers in plain English, no SQL needed. Marketing costs fell 70%.
- Deployed a RAG support chatbot on Snowflake Cortex that answers from internal docs, using hybrid retrieval and reranking. Hit ~95% Recall@5 and ~90% Precision@5.
- Provisioned AWS infrastructure (S3, Lambda, Glue, Kinesis, MSK) with Terraform and rebuilt batch jobs as serverless microservices. Runtime went from 3+ hours to under 15 minutes — 10x faster.
- Engineered batch and streaming ETL/ELT pipelines from PostgreSQL, MongoDB, Kafka (AWS MSK), Kinesis, and APIs into Snowflake's Medallion layers, then tuned them: 70%+ faster, 5x the data, 40%+ cheaper.
- Led 4+ engineers as Primary Consultant, turning client problems into working AI and data products from idea to launch.
- May 2023—Jul 2023Internship
Data Science Intern
UIDAI Technology Centre·Bangalore
- Trained and evaluated ML and deep-learning models for biometric fraud detection.
- Ran CNN-based fingerprint experiments to sharpen anomaly detection, tracking precision, recall, and false-acceptance rate.
- Wrote end-to-end ML workflows in Python with TensorFlow, OpenCV, and scikit-learn.
- Dec 2022—Aug 2023Internship
Research Intern
Samsung Research Institute·Remote · Bangalore
- Developed deep-learning pipelines for pose estimation and keypoint detection using CNN models.
- Created annotation, labeling, and validation tooling for the research team.
- Turned experimental research models into reusable ML components, and cleaned training data to reduce bias.
Four areas, one practice.
The shape of my day: shipping AI agents and RAG, and building the data foundation that keeps them grounded — pulling data, modeling it, and making the whole thing safe to change.
01
Agentic AI & LLM Engineering
Production AI agents, RAG, and text-to-SQL.
- Agentic workflows and autonomous AI agents on the Claude API (Anthropic)
- Retrieval-Augmented Generation (RAG) and text-to-SQL over enterprise data
- Snowflake Cortex agents, AI-ready semantic layers, prompt orchestration
02
Snowflake
Warehousing architecture and AI-driven analytics.
- Data warehousing architecture (Medallion Architecture)
- ELT pipelines for batch and real-time loads — Snowpipe, Streams, Tasks, Dynamic Tables
- Cortex agents, semantic layer, and Snowflake Intelligence
03
Data Engineering
Pipelines built to run reliably in production.
- Batch and streaming ETL/ELT from Postgres, MongoDB, Kafka (MSK), Kinesis, and APIs
- Star-schema modeling, data quality checks, and reconciliation
- Python and SQL orchestration for cloud-native pipelines
04
Cloud & DevOps
AWS data stack, infrastructure-as-code, and CI/CD.
- AWS data stack — S3, Lambda, Glue, Step Functions, Kinesis, MSK
- Terraform infrastructure-as-code and serverless, microservices design
- CI/CD with GitHub Actions and Azure DevOps
The toolbelt
Connect with me.
Recruiter, founder, collaborator, or just curious — drop a note and it lands directly with me.