Solutions engineer, forward deployed
I own the whole arc, from the first discovery session to go-live. LLM agents clearing 77% of licence renewals for 1.5M+ practitioners, a judicial case-management platform made 70% faster, a 10M-record database migration, and before that, co-founder and CTO of an AI startup that reached 15,000 users.
Government and healthcare work via Ignyte Group
Simplified illustrations of real work. Press play, or step through.
FIG 1 / Florida Department of Health
Automate routine licence renewals for 1.5M+ practitioners without ever letting a model deny one.
77%of renewals auto-processed
Open case studyLicence renewals are split into fields, checked in parallel by an LLM agent and a business-rules layer, and routed to either auto-approval or human review. The agent has no way to deny.
FIG 2 / Massachusetts Administration and Finance
Pages were slow across a case system for four judicial appeals boards, and for more than one reason.
A page load waterfall shows repeated Microsoft Graph calls and an unindexed dashboard query. Batching the calls and moving the query to an indexed view cuts page load time by 70%.
70%less page load time
Open case studyFIG 3 / Networkqy AI
Users asked the same questions in different words, and every one was a paid API call.
A retrieval pipeline answers a new question through Pinecone, a cross-encoder and a paid API. A reworded repeat of the question is matched by a semantic cache and answered by a self-hosted model, skipping the paid call.
Repeat questions served by a self-hosted model, not the paid API.
Open case study01
12+ regulator sessions
02
112 traced requirements
03
20+ RFP sections, 2 winning bids
04
LLM agents in production
05
30 staff trained
What I got wrong first
Working solo. I used to think owning every piece myself was the fastest way to ship. The best work on this page came from teams: regulators who knew the rules inside out, developers who caught what I missed, and caseworkers who told us what actually broke. Now I build the team’s shared understanding before I build the system.
Most of my work sits where software meets regulation: licensing, appeals boards and health services. I like starting at the shadowing sessions and staying until the system is stable in production.
Full history on LinkedInForward Deployed Engineer (Technology Consultant)
Apr 2025 – Present
Massachusetts A&F, Legal Case Management System
Apr 2026 – Present
Building the case lifecycle engine that moves four judicial appeals boards onto one Appian platform: state transitions, parent-child case consolidation, and multi-day hearing scheduling. Traced app-wide latency to Graph API fan-out and unindexed dashboard queries, then fixed it with request batching and indexed SQL Server views. Owned the Civil Service Commission go-live.
Stack: Appian · Java · SQL Server · AWS S3 + Lambda · Microsoft Graph API · OAuth 2.0
Florida Department of Health, MQA Licensing System
Apr 2025 – Mar 2026
Ran 12+ discovery sessions with regulators, then architected renewal decisioning on Claude via AWS Bedrock: a JSON-schema decision contract, a parallel validation layer, and no denial without human review. Built the integrations around it, including Tyler payments with settlement reconciliation, CE Broker, AHCA, and SAML SSO on the state’s Entra ID.
Stack: Appian AI Agents · Claude on Bedrock · REST APIs · SAML SSO · SQL stored procedures · Selenium
Associate Forward Deployed Engineer (Associate Consultant)
Feb 2024 – Apr 2025
Rocky Mountain Human Services, Case-management expansion
Feb 2024 – Mar 2025
Ran day-to-day delivery for the client’s five-person dev team and built a custom Java plugin. Led the MySQL to SQL Server migration with UTF-8 hardening and walked stakeholders through cutover risk before go-live.
Stack: Java · MySQL · SQL Server · Appian
Software Developer
Aug 2023 – Dec 2023
Replaced manual Excel reporting with a Google Tag Manager to Tableau analytics pipeline, and shipped Vue.js/Nuxt features to kodak.com behind a containerised Git and Docker release workflow.
Stack: Vue.js · Nuxt · Docker · Craft CMS · Tableau
Co-founder and CTO
Apr 2021 – Jul 2023
Co-founded an angel-funded career platform for universities and built the AI stack end to end: real-time interview practice on Whisper and GPT-4 over WebSockets, a Pinecone RAG pipeline with cross-encoder reranking, and a semantic cache on a self-hosted, LoRA-tuned Mistral 7B. Ran every pilot from first call to paid contract.
Stack: Python · React/Next.js · Pinecone · vLLM · OpenAI · MongoDB
Business Technology Intern
May 2023 – Jul 2023
Moved an Excel-only master data team onto SQL, Python and Power Platform, adding pipeline monitoring and event-driven alerts against SAP and Azure SQL.
Stack: SQL · Python · Power BI · Power Automate
MS Computer Science, Data Science
University of Southern California, 2023
BE Computer Science, minor in Finance
BITS Pilani, Dubai, 2021
NETWORKQY AI
Whisper speech-to-text chained into GPT-4 inference, streamed over WebSockets so mock interviews run in real time.
Whisper · GPT-4 · WebSockets
NETWORKQY AI
RAG over users, skills, companies and jobs: 256-token chunks, top-5 retrieval and cross-encoder reranking, scored on precision and recall.
Python · Pinecone · LangChain
NETWORKQY AI
A LoRA fine-tuned Mistral 7B, self-hosted on vLLM, matching repeat queries and answering them locally instead of calling OpenAI.
Mistral 7B · LoRA · vLLM
USC COURSEWORK, CODE PRIVATE
Processes, threads, a virtual file system and virtual memory for a teaching operating system.
C · QEMU · GDB
I’d like to join your team. sachinsingh018@gmail.com