Currently focused on AI Engineering and emerging technology. I'm driven by purpose, creativity, and a desire to push myself into experiences that expand how I see the world.
I'm at my best when I'm learning fast, keeping my imagination sharp, building something impactful, and moving toward the edge of what feels comfortable.
I'm a Senior Solutions Engineer focused on applied AI, cloud architecture, and turning complex customer problems into systems that actually work.
Over the past six years, I've worked across cloud infrastructure, APIs, data, Kubernetes, and enterprise technical evaluations at VMware, Broadcom, and CloudZero. More recently, I've been building AI systems myself, mostly around agents, evaluation frameworks, retrieval workflows, developer tooling, and automation designed around real-world problems.
What interests me most is the space between product, engineering, and the customer: understanding the problem deeply, building the right solution, testing where it breaks, and turning what works into something repeatable.
Technical pre-sales at a cloud cost intelligence platform managing $14B+ in customer cloud spend. Run POCs, deliver technical demos, and help engineering and finance teams build cost allocation models across AWS, GCP, and Azure.
Cloud management, security, and application observability post-VMware acquisition. Maintained technical relationships and pipeline during a complex org transition.
Led technical evaluations for VMware's cloud cost management and security platforms. Built the Technical Enablement Program, closed the largest CloudHealth commercial deal, and maintained 90%+ customer renewal rates.
Owned post-sale success for 100+ customers. Built onboarding programs and automated outreach workflows that drove consistent retention above 90%.
Grew pipeline through prospecting and discovery. Entry point into cloud cost management and the SE track.
Show More
- Turned AI into the reason to buy. Demonstrated per-user, per-model, and per-department cost for Claude, Copilot, ChatGPT, Gemini, Cursor, Bedrock, and Vertex AI to a data team hand-building the same pipelines. They became the champions who pushed procurement to sign early.
- Cleared the FedRAMP blocker in week three. Wrote the script that mapped Azure GovCloud billing into CloudZero's ingestion format, proving compliant coverage before the customer had to build anything.
- Automated a full-time manual process. Built the dimensions and telemetry splits that allocate 11 AWS billing organizations of shared spend back to customers, replacing a monthly pipeline one engineer ran by hand.
- Found real money fast. ~$200K/year in orphaned KMS keys, a $67K storage spike, and a $44K database anomaly, all surfaced within days of connecting data.
- Gave the CIO board-ready numbers. Built the drill-downs and reservation coverage views used to negotiate $500M+ in five-year cloud commitments.
- Built champions in six teams. Cloud ops, finance, procurement, engineering, the AI data team, and the CIO office each had their own reason to want the platform, so the deal survived a CFO change and a months-long stall.
Show More
- Scoped a proof-of-concept spanning AWS, Snowflake, 8 EKS clusters, and CI-pipeline spend brought in through CloudZero's bring-your-own-data ingestion.
- Walked the evaluator through the Kubernetes cost algorithm, max(request, usage) per pod, and proved out no node-level double-counting, plus PVC/EBS reconciliation and reservation pass-through.
- Worked a hard edge case: attributing network egress for blockchain validator traffic at flow-level granularity.
- Resolved an SSO configuration lockout before the final close call, and built engineering dashboards cost-ranked to stay inside the BI layer's row limits.
Show More
- Scoped the trial around explicit objectives: AWS + Kubernetes ingestion, product/team/customer allocation, and quantified ROI, then imported a year of historical billing data to prove trends, not snapshots.
- Modeled shared-service allocation using proportional usage-based splits, and answered the methodology questions head-on: reservation coverage, anomaly detection, and how shared platform costs flow to products fairly.
- Recorded short async video walkthroughs (cost-splitting explained with an airplane-seat analogy) so executives who missed live sessions could still follow the business case.
- Survived a champion departure mid-evaluation by having already built relationships across 20+ stakeholders, from engineers to the CFO.
Show More
- Connected AWS, Databricks, and Snowflake billing in the first session, then built "data domain" dimensions from the customer's crosswalk files so every Databricks job ID rolled up to the end-client it serves.
- Cleaned up account-naming inconsistencies blocking allocation, and ran a structured financial-dashboard review with the finance team against their real reporting needs.
- Scoped API/IaC automation so onboarding each future client wouldn't require touching the cost model by hand.
- After close, the CISO brought third-party pricing changes to Griffin for impact analysis: pre-sales trust converting into an advisory relationship.
Show More
- Designed the CAD/USD conversion as monthly FX-rate telemetry uploads, so finance reporting landed in the currency the business actually budgets in.
- Split shared Cassandra costs by what drives them: storage allocated by event volume, compute by read/write activity, and GKE workloads by app labels with shared cluster costs distributed on CPU/memory.
- Brought GCP discount mechanics (committed-use and sustained-use) and BigQuery reservations into the allocation model rather than leaving them as unallocated noise.
- Configured RBAC and scoped views so sensitive cost data stayed hidden from the wrong audiences (the failure that had sunk their previous Kubernetes cost tool), plus budget alerts delivered into Slack.
Show More
- Diagnosed in the first demo that per-customer RDS attribution (~$130K/month of database spend) was the make-or-break requirement, and outlined a custom telemetry path before the POV started.
- Connected AWS plus dev and prod EKS clusters live during the kickoff call, with compute allocation keyed off pod labels.
- Validated the table-to-customer telemetry workflow in a working session against their real data, including a clean CSV contract so their team could produce the feed reliably.
- Designed the maintenance path as API-driven, so new customer onboardings flow into the cost model automatically instead of accruing manual work.
Show More
- Automated multi-source billing ingestion where the platform's standard connectors didn't cover the customer's estate.
- Built a cost allocation model complex enough to be the deal's core technical risk, and de-risked it during the evaluation rather than after signature.
- Part of a run of 100%+ quota across 12 consecutive quarters, with 150%+ in 3 of them.
Show More
- Built the reusable assets the motion runs on: demo environments and dashboard templates for cloud unit economics, cost allocation, and Kubernetes spend.
- Wrote Python data-ingestion pipelines (AWS Lambda, S3, REST APIs) to automate cost allocation and margin reporting for evaluations.
- Owns security and compliance as first-line technical contact: RFPs, security questionnaires, and buyer questions on access controls and data handling.
- Coaches new SEs and AEs on discovery, POV execution, objection handling, and demo strategy.
Show More
- Success criteria are written and agreed with engineering and finance stakeholders before any data is connected, so "done" is defined before the clock starts.
- Validations run against the customer's real billing data, structured as working sessions that build the actual allocation model rather than a throwaway demo.
- Ends in a readout that maps each criterion to evidence, which is what compressed 30 days to 2 weeks without cutting rigor.
Show More
- A generate → validate → repair loop: agents must run a ~1,000-line Python validator (16 deterministic lint rules) and fix errors before presenting YAML. The architecture assumes the model will sometimes be wrong and catches it.
- A routing table sends agents to exactly the corpus files a task needs, and a mandatory pre-generation checklist makes compliance observable to the human reviewer.
- 20 worked example patterns (anonymized from real customer configs) with selection metadata, plus an 18-case eval harness including negative tests that verify broken YAML gets caught.
- One corpus shipped in four agent-native formats (CLAUDE.md, .cursorrules, Copilot instructions, AGENTS.md), kept isomorphic so behavior is identical across tools. Public at github.com/griffinsisk/costformation-brain.
Show More
- Structured as recurring cross-training where each product's SEs taught the others their demo, discovery questions, and qualifying signals.
- The payoff: any SE could spot and tee up opportunities for the whole portfolio, instead of selling one product into accounts that needed three.
Show More
- Used customers' own cost data as the trigger: spend patterns and feature-adoption gaps generated outreach with a concrete reason to talk, not a check-in.
- Paired with a scaled NPS program across a 100+ account book, together driving renewal above 90%.
A few things I've designed and shipped. I care as much about the context design, evaluation, and reliability behind these as the prompts themselves.
A serverless relay that reads the AWS China cost and usage report, rewrites every money column from yuan to US dollars while keeping the report's exact format, and copies it to a commercial-partition bucket so CloudZero ingests it as native AWS data.
- Idempotent by design: a deterministic ID built from the source version and exchange rate means a month is only reprocessed when the data or the rate changes.
- Exact decimal arithmetic for money, verified by a round-trip test that must match the original to the cent.
A routed knowledge corpus, a custom linter with 131 tests, 20 worked example patterns, and an eval harness, packaged so Claude Code, Cursor, Copilot, and Codex all behave the same way.
- Constrain and verify instead of trusting generation: the linter catches the model when it is wrong.
- Observable compliance: agents must state which files they read before generating.
- One corpus shipped in four agent formats, kept identical so behavior matches across tools.
A grounded ElevenLabs agent that answers from a source-tagged profile, takes actions through client tools such as showing projects and opening a booking page, plus a bring-your-own-AI onboarding kit and a validator so anyone can build their own.
- Every fact carries a source tag, and a validator blocks untagged claims, phone numbers, and placeholder text before anything ships.
- Tools for actions instead of generated text: links and bookings are deterministic.
- Voice runs over WebRTC and chat over WebSocket through one agent, under a strict content security policy.
A Chrome extension that logs each click and navigation during a demo, plus a Python pipeline that describes each step with Claude vision, fact-checks a written narration, voices it with ElevenLabs, and assembles the final video with captions.
- Claim-level verification of the narration against frames and cited sources before any audio is generated.
- A human approval gate before every paid step, and resumable runs that only redo stages whose inputs changed.
A LLM Assistant that guides a proof-of-value from success criteria to a validated business case, with state kept in shared storage so any sales engineer can pick it up. It cuts each evaluation by more than two hours.
- Synthesizes customer calls into prioritized tasks against SOW success criteria, using focused RAG, multiple MCPs, and Python validation logic.
- Automates POV stand-up and cost allocation, cutting time-to-value by ~2 hours per POV.
A two-pass culling app with intent-aware grading, taste profiles, and editorial guidance, with billing and a bring-your-own-key option.
- Downscaling for the cull pass cut vision-token cost about 60 percent.
- Measured determinism instead of assuming it: 340 API calls showed 100 percent rating stability at temperature 0.
- Regression-tested scoring with eval fixtures on every prompt change.
An event-driven serverless pipeline plus a backfill tool, built for customers to deploy and run themselves.
- Selective retries with exponential backoff, and structured JSON logging for supportability.
A set of AI agent skills that each handle one stage, from assessing fit to preparing for an interview, with evidence tracking and a human approval required between stages.
- Human approval is the boundary between stages; nothing is sent automatically.
An app that runs locally, counts what a cleanup would remove before acting, and trashes mail in batches of 1,000 with undo.
View on GitHub ↗Echo is my AI assistant, grounded in a reviewed profile of my work. It says plainly when something isn't covered, and can point you to projects and stories on this page. It's one of my projects: ask Echo how it works.
Try asking out loud
Prefer to type?
AI assistant built with ElevenLabs. Conversations are processed by ElevenLabs and its model providers.