Commercial AI
AI customer-operations SaaS
Built and led an AI customer-operations platform to $1M ARR; shipped autonomous agents that improved customer resolution rates by an average of 30%.
$1M ARR · 30% average resolution improvement
Senior AI delivery · production systems · measurable outcomes
Production AI that reaches revenue and operations.
MahumTech brings senior product, architecture, data, cloud, and engineering leadership to AI programs—from diligence and recovery through secure deployment, operating evidence, and handover.
- Senior delivery team
- International client delivery
- Evidence-gated releases
$1M ARR· AI customer-operations SaaS
+30%· average resolution improvement
50 locations· HIPAA pharmacy analytics live
CrewScore + GitPin· released open source
Anonymized client systems
Case evidence across commercial and regulated AI
These engagements show the operating problem, delivered system, and verified boundary or outcome. Client names, product names, domains, repositories, contacts, and distinctive internal labels remain private.
9 selected engagements · identities withheld
Regulated healthcare
Regulated pharmacy analytics
Delivered HIPAA-compliant NLP-to-SQL analytics live across 50 pharmacy locations and connected medication-management devices for an approximately $100M enterprise.
Live across 50 pharmacy locations
Commerce operations
Commerce support automation
Built a multi-intent support agent for a $1M ARR Shopify portfolio, integrating email, commerce, shipment tracking, knowledge rules, policy gates, evaluation, and human escalation.
$1M ARR portfolio · human-gated automation
Private legal AI
Private legal intelligence
Built private, on-premises legal-intelligence workflows in a client-controlled environment, with governed retrieval, drafting assistance, and human review.
On premises · client-controlled data boundary
Enterprise integration
Enterprise AI API and identity
Delivered an enterprise AI API secured with Entra ID, OAuth2/JWT, tenant and user access controls, strict JSON contracts, and production Azure deployment.
Identity-gated · production Azure deployment
Financial services
Corporate credit-risk automation
Built an agentic workflow for evidence gathering, structured case analysis, risk and complexity assessment, collection mandates, and expert review.
Cross-border decision workflow · human validation
Knowledge work
Technical-document AI
Delivered AI-assisted technical-document generation with enterprise search, section-aware assistance, structured editing, templates, and workflow integration.
Private knowledge workspace · structured authoring
Conversational operations
Conversational lead qualification
Delivered multi-step voice and messaging qualification, scheduling, and live-call transfer with monitoring, runbooks, security controls, and operator handoff.
Live human transfer · operational handover
Executive technology leadership
Consumer-health interim CTO
Led a three-month interim-CTO engagement through platform stabilization, delivery governance, architecture, security and privacy controls, and beta readiness.
Three-month stabilization · beta-readiness handover
This record includes Pendoah-delivered and independent confidential engagements led by MahumTech's founder. They are presented as leadership and delivery evidence; no engagement is relabeled as a MahumTech contract.
Operating model
Production AI with evidence at every boundary
We design the full operating system around the model: context, execution, tools, evidence, approvals, and release discipline.
Bound context and decisions
Context engineering starts with the workflow, source authority, user, and decision boundary—not an unbounded chat interface.
Engineer durable execution
Long-running agent work uses explicit state, retries, idempotency, checkpoints, and resumable human handoffs.
Gate releases with evidence
Deterministic tests, groundedness and citation checks, adversarial evals, and revision-pinned proof turn evaluation into a release gate.
Keep agency governed
Least-privilege tool scopes, audit trails, and human approval keep consequential side effects outside the model’s authority.
Senior team delivery
Architecture, execution, and accountability stay connected
MahumTech works with senior operators who can move between product strategy, system design, implementation detail, delivery recovery, and executive decision-making without losing the evidence trail.
Engagements are shaped around the decision and operating constraint, with explicit ownership, measurable acceptance evidence, and a handover path from the beginning.
Product and executive leadership
Senior operators stay accountable
Technology strategy, product decisions, architecture, security, and delivery governance stay connected from diligence through production handover.
Principal architecture
The hard boundaries are designed first
Identity, tenancy, data authority, retrieval, tool permissions, evaluation, and human approvals are treated as system architecture—not prompt decoration.
Engineering and platform delivery
Teams carry the system into operation
Senior product, application, data, cloud, and AI engineers work across the full stack, including tests, deployment evidence, observability, and runbooks.
Regulated and high-consequence work
Control survives contact with production
Least-privilege access, auditable queries, source grounding, exception paths, and explicit review gates support healthcare, legal, financial, and enterprise workflows.
Production agent systems
Architecture, release evidence, and governance move together
Agent runtime and context
Multi-agent orchestration and durable execution pair structured state with context engineering, hybrid RAG, reranking, citations, and bounded memory.
Interoperability by need
MCP tool surfaces, authenticated APIs, and scoped agent interoperability are introduced only when cross-system exchange is justified; otherwise the surface stays closed.
Evaluation and observability
Release gates combine deterministic and adversarial evals with GenAI observability for model, retrieval, tool, cost, latency, and approval evidence.
Governance that can be examined
Least-privilege access and HITL controls map to NIST AI RMF and OWASP GenAI and agentic risks; ISO/IEC 42001 is treated as readiness work, not a certification claim.
Leadership
Sarosh Hussain
Founder, MahumTech · CTO & Partner, Pendoah
Commercial AI CTO and former enterprise CIO who leads senior client delivery while remaining hands-on at architecture, security, evaluation, and production-release boundaries.
Public open-source proof
Released tools with inspectable boundaries
CrewScore and GitPin show the same control, evidence, and release disciplines used in client delivery. Both are public and released; no commercial traction, revenue, or adoption is implied.
2 open-source releases
AI agent guardrails
CrewScore
Released open source
crewscore.ai
An offline deterministic check for 23 published controls across eight dimensions—not proof of runtime safety.
Agent delivery assurance
GitPin
Released open source
shmindmaster.github.io/gitpin
GitPin v0.6.3 is a read-only, Git-HEAD-pinned search → prove → verify gate with re-checkable path, line, and commit evidence.
Open-source adoption
CrewScore and GitPin are available now
Start with the published controls and verification boundaries, then decide whether either tool belongs in your agent-delivery workflow.
Choose the engagement path that matches the work
MahumTech separates production AI delivery, architecture and interim leadership, and open-source adoption so every conversation starts with a clear boundary.
Production AI delivery
Bring a consequential workflow
Share the operating constraint, source systems, decision owner, and the outcome that must hold in production.
Discuss a delivery engagementAvailable now
Evaluate released tools
Review CrewScore or GitPin, their published boundaries, and the adoption path before deciding whether either fits.
Evaluate the open-source toolsArchitecture and interim leadership
Stabilize a product or delivery system
Use a bounded senior-led engagement for architecture diligence, production recovery, governance, or beta readiness.
Discuss the operating gapSchedule with MahumTech
Prefer async or phone? Use the form or call +1 346-420-5659.
Start with the operating outcome that matters most
Bring a production AI program, architecture or delivery recovery need, product collaboration, or open-source adoption decision.
Share your inquiryContact
Start with fit, not a sales package
Tell us the workflow, constraint, decision owner, and intended outcome. Your message goes directly to the MahumTech team for a bounded next-step decision.
- 1
Choose the path
Name the operating outcome, product, or architecture boundary you need to address.
- 2
Make the boundary explicit
Share the operating constraint, decision owner, and what must remain human-reviewed.
- 3
Leave with a next step
Continue to a scoped delivery conversation, an open-source evaluation path, or a clear no-fit.
Talk with MahumTech
Delivery and product inquiries
Use the form or email [email protected].
Product support
[email protected]Phone
+1 346-420-5659Office
6847 Emerson Ln, Sugar Land, TX 77479
Prefer a scheduled conversation?
Choose a time on Calendly. Opening the scheduling page does not create an appointment until you confirm it there.
Open CalendlyRead our Terms and Privacy Policy.