h
himansh_srivast

Himansh

@himansh_srivast

AI and Automation Engineer

Estados Unidos
Inglés, Hindi
Parte de la información aparece en idioma inglés.
Sobre mí
I build Python automation, AI integrations and data pipelines for teams stuck repeating the same work by hand every week. I am an AI engineer at a regulated mortgage lender. One process I automated there replaced 2 to 3 hours of a specialist's day. I have 5 releases on PyPI, including a signed audit trail for AI decisions, and I have trained and shipped models in PyTorch and TensorFlow. Python, LLM and RAG integration, OpenAI and Claude APIs, SQL, ETL, Airflow, PySpark, Power BI. I reply within a few hours, and I tell you before you order if I am not the right fit.... Lee más

Habilidades

h
himansh_srivast
Himansh
desconectado • 
Tiempo medio de respuesta: 14 horas

Revisa mis servicios

Automatizaciones
I will automate your repetitive excel and data work with a python script
DeUSD 50 / proyectoMás información
Banners y publicaciones para redes sociales
I will design on brand social media posts and editable canva templates
DeUSD 15 / proyectoMás información

Porfolio

Experiencia laboral

Mortgage_Loan Center

Ai Engineer intern

Mortgage Loan Center • Tiempo parcial

Jun 2026 - Present • 4 mos

Production data and AI systems inside a regulated mortgage lender, where the output is read by auditors and regulators. GL reconciliation. A specialist spent 2 to 3 hours a day cross-referencing Power BI, the loan origination system and the A&B ledger to classify and route mortgage GL discrepancies. None of it was written down, so I ran structured interviews and sat through live sessions, then codified 16 rules that could be tested. It now routes 100+ open cases by debit and credit class across 3 joined systems. Ginnie Mae pool delivery. ULDD XML exports from 3 origination systems (Encompass, MeridianLink, Blue Sage) merged into valid Ginnie Mae PDD XML for SFPDM submission, with MISMO schema errors corrected in place rather than bounced back to a person. First deployment covered $18.2M UPB across 4 pools and 47 loans. CI/CD security. Built the security stage of a five-stage deployment platform, on the rule that a pipeline cannot grant itself permissions. Alongside this I published custody-ledger on PyPI: a signed hash chain that proves what a model produced and what a human did about it, with 151 tests running against 2 database backends. Skills field — type short prefixes, Fiverr forces you into its taxonomy: Python · SQL · ETL · Data engineering · Power BI Machine learning · Automation · XML · CI/CD · Data analysis Industry Financial Services Every number traces to your own site: 2-3 hrs daily, 16 rules, 100+ cases, 3 systems, $18.2M UPB, 4 pools, 47 loans, 151 tests, 2 backends. Nothing invented. The only thing I can't verify is your start date, so fill that yourself. Two notes on the writing. I opened with the regulated-lender context because that's the part a buyer can't get from a cheaper freelancer, and I kept the "pipeline cannot grant itself permissions" line because it's the most memorable sentence on your whole portfolio and it signals judgment rather than just tooling.

Syracuse_University

AI/ML Intern

Syracuse University

Aug 2023 - May 2025 • 1 yr 9 mos

Reproduced SimCLR contrastive learning and applied it to a new image dataset in PyTorch and TensorFlow. Built ETL pipelines in PySpark and Airflow over 1M+ records.