Uttam PatelAI Engineer & Data Scientist

generate(identity) →

AI Engineer & Data Scientist

I build AI that ships.

  • ML modelsPredictive modelling, deep learning and ensembles that hold up in production.
  • LLM agentsAgents, AI workflows and n8n + Gemini automations that do real work.
  • SaaS productsFull-stack products, from data model to deployed app, e.g. CliniqEase on Django.
  • Data systemsPipelines, forecasting and quant systems, including trading on Indian markets.
View the models Skip to work

From idea to production.

Machine learning models, LLM agents and automations, SaaS products and data-driven systems, taken from idea to production.

I hold a B.Tech in Computer Science and work across data science, machine learning, deep learning and applied LLMs.

A web development background sits alongside the data work, so what I build ends up in products people use: fast, clear and deployed. I also run a web & digital agency and build my own SaaS.

role
AI Engineer & Data Scientist
ships
ML models · LLM agents · SaaS · data systems
founder
Hunexture (agency) · CliniqEase (SaaS)
education
B.Tech in Computer Science
based
Ahmedabad, India
contact
utampipaliya@gmail.com

  1. Machine learning

    Classification, regression and clustering. Predictive analytics, feature engineering and tuning.

  2. Deep learning

    CNNs, LSTMs and Transformers in PyTorch and TensorFlow, for vision, language and sequences.

  3. LLMs & RAG

    Retrieval over your own documents, chat interfaces, fine-tuning with LoRA / QLoRA when it pays off.

  4. Agents & automation

    LLM agents and n8n + Gemini pipelines that take repetitive work off a team’s plate.

  5. Forecasting & quant

    Time series forecasting, anomaly detection and algorithmic trading systems.

  6. Data & dashboards

    Statistical analysis, interactive dashboards and real-time monitoring.

  7. SaaS & full-stack

    Django, React and Node products, built end to end and put in front of users.

  8. MLOps & deployment

    Pipelines, versioning, containers and cloud deployment with monitoring.

  9. Big data

    Kafka and Spark pipelines for large-scale and real-time processing.

Models in production.

Every project is a model with a card. Hover one, or press Run inference, to read its stack, status and one real metric.

df.sort_values("featured") → 28 rows

  • LLM

    Multi-agent task automation

    Planner, Researcher, Writer and Reviewer agents turn one goal into a cited brief, with tool use, a bounded revision loop, a grounding audit and a claim-level fact-checker.

    4agents in the loop, every step auditable
    > inference(model="multi-agent")
STACK    Gemini · FastAPI · Pydantic · Docker · GitHub Actions
STATUS   open source
METRIC   4 agents in the loop, every step auditable
CLASS    LLM Agents & Automation
  • LLM

    RAG knowledge assistant

    Document Q&A grounded in your own PDFs, with inline citations. Hybrid dense + BM25 retrieval fused with Reciprocal Rank Fusion, optional MMR re-ranking.

    28pytest tests across retrieval and pipeline
    > inference(model="rag-assistant")
STACK    Gemini · sentence-transformers · FastAPI · Streamlit · pypdf
STATUS   open source
METRIC   28 pytest tests across retrieval and pipeline
CLASS    LLM Agents & Automation
  • SaaS

    CliniqEase

    Clinic management SaaS: the day-to-day running of a clinic in one product. Built with Django.

    > inference(model="cliniqease")
STACK    Django · Python
ROLE     Founder
CLASS    SaaS Products
  • Quant

    NSE trading strategy backtester

    A vectorised backtester for Indian large-caps: trend and mean-reversion rules against buy-and-hold, net of trading costs, with walk-forward optimisation to avoid look-ahead bias.

    −11.5%max drawdown vs −33.8% buy-and-hold (backtest, synthetic data)
    > inference(model="nse-backtester")
STACK    Python · pandas · NumPy · yfinance · Streamlit
STATUS   open source
METRIC   −11.5% max drawdown vs −33.8% buy-and-hold (backtest, synthetic data)
CLASS    Quant & Data Systems
  • Web3

    Blockchain academic credential verification

    Academic credentials issued and verified on a blockchain, so a degree can be checked without calling the institution.

    > inference(model="credential-chain")
STACK    Blockchain
CLASS    Web3
  • Agency

    Hunexture

    My web & digital agency: websites and digital products for clients, designed, built and shipped.

    > inference(model="hunexture")
ROLE     Founder
CLASS    Agency & Client Work

The toolkit, embedded.

Skills as points in a drifting 3D space. Hover or focus a cluster to isolate it; hover or focus a skill to see its nearest neighbours.

The training log.

Experience as a training run: each role is an epoch, each milestone a checkpoint.
Illustrative curve · the log on the right is the real record

  1. [EPOCH 01/04]—

    B.Tech, Computer Science

    • Data science, machine learning and deep learning foundations
    • Web development alongside the data work

  2. [EPOCH 02/04]—

    Founder · Hunexture

    • Started a web & digital agency
    • Client websites and digital products, designed, built and shipped

  3. [EPOCH 03/04]—

    Founder · CliniqEase

    • Built a clinic management SaaS on Django

  4. [EPOCH 04/04]—

    AI Engineer & Data Scientist

    • Machine learning models, LLM agents and data systems taken to production

Held-out questions.

q_00What do you build?

Machine learning models, LLM agents and automations, SaaS products and data systems. Usually the whole path: data, model, the app around it, and deployment.

q_01What is your expertise in deep learning?

CNNs for images, LSTMs and Transformers for sequences and text, graph neural networks and self-supervised learning. Mainly PyTorch and TensorFlow, with compression and distributed training when models need to run fast or at scale.

q_02How do you approach generative AI projects?

From the task and the data, not the model. Most projects begin with retrieval (RAG) over your own documents, move to fine-tuning with LoRA or QLoRA only when prompting and retrieval aren’t enough, and include an evaluation set from day one so quality can be measured.

q_03Do you take on client and agency work?

Yes. Through Hunexture, my web & digital agency, and directly for AI and data projects. The contact form at the end of the page is the fastest route.

q_04How do you make sure a model holds up after launch?

Clean validation splits (time-based for time series), metrics chosen for the business problem rather than accuracy alone, error analysis on the cases the model gets wrong, and monitoring for drift once it is deployed.

Send a request.

Building something with AI, data or a product that needs to ship? Send the request; a human answers.

utampipaliya@gmail.com
Near L.D. Engineering College, University Road, Ahmedabad, Gujarat 380006
POST/collaborate