Production AI systems architected and delivered for enterprise clients.
Clients are anonymized, but the outcomes are real.
AlixPartners · Private equity clients
Agentic M&A Due Diligence Platform
ProblemM&A due diligence
means analysts manually going through data rooms, like headcount
files, vendor lists, and financials, under tight deal deadlines.
BuiltArchitected and shipped a
Generative AI platform where agents run agentic RAG workflows:
multi-step reasoning, planning, and tool execution across headcount
analysis, vendor mapping, and peer benchmarking.
ImpactIn production with
private equity clients, automating diligence work that used to be
fully manual.
Agentic RAG · Multi-step planning · Tool execution
AlixPartners · Platform infrastructure
MCP Integrations for Enterprise Agents
ProblemAgents needed access to
a growing set of internal and external data sources and services.
Each one needed custom integration code that was hard to standardize
or reuse.
BuiltBuilt Model Context
Protocol (MCP) integrations connecting AI agents to enterprise data
sources and services, giving every agent one standard way to
discover and call tools.
ImpactStandardized tool use
across the due diligence platform. New data sources plug in through
one protocol instead of one-off integrations.
MCP · Tool / function calling · Enterprise APIs
AlixPartners · Procurement org, $3.5B annual spend
GenAI Contract Lifecycle Platform
ProblemProcurement teams
reviewed contracts by hand: classification, clause-by-clause
reading, and redlining. Each cycle took weeks.
BuiltDesigned and shipped a
contract lifecycle management platform; owned prompt and system
design, retrieval, and extraction quality, iterating on evaluations
in production across contract classification, clause extraction, and
AI-driven redlining.
ImpactCut contract review
cycles from weeks to hours for a client with $3.5B in annual
spend.
Retrieval · Production evals · AI redlining
AlixPartners · Customer service operations
LLM Call-Analytics Application
ProblemRecurring service
issues were buried in customer calls that no team had the capacity
to listen to and categorize.
BuiltPrototyped and
productionized a speech-to-text call-analytics application, using
LLMs to surface and group recurring service issues from transcribed
customer calls.
ImpactTurned raw call audio
into an actionable issue feed, driving customer satisfaction
improvements.
Speech-to-text · LLM classification · Prototype → production
AlixPartners · National restaurant chain
ML Store-Performance Program
ProblemWhich stores
underperform, what a closure actually does to revenue, and what the
franchise contracts actually say. Three questions that used to be
answered by gut feel.
BuiltDelivered an ML program
combining classification models for store performance, causal
modeling of closure impact on revenue, and LLM-based extraction of
terms from franchise contracts.
ImpactStore performance and
closure decisions backed by models and extracted contract terms
instead of intuition.
Classification · Causal inference · LLM extraction
Sia Partners · Enterprise search
Domain-Tuned Retrieval Backend
ProblemOff-the-shelf
embeddings missed domain vocabulary, and AI answers without source
pointers couldn't be audited.
BuiltFine-tuned a
domain-specific embedding model with QLoRA and built a large-scale
retrieval backend on Elasticsearch and Pinecone, with page-level
metadata attached to every indexed chunk.
ImpactEvery AI answer stayed
traceable to its source page, giving auditable retrieval at scale.
QLoRA fine-tuning · Elasticsearch · Pinecone
Sia Partners · ESG reporting
ESG Compliance Platform
ProblemCompliance work
depended on large volumes of unstructured ESG reports that had to be
read and structured manually.
BuiltBuilt a cloud-based ESG
compliance platform with a scalable OCR pipeline on AWS Textract,
automating extraction and structuring of unstructured reports.
ImpactManual document
processing replaced with an automated pipeline that scales with
report volume.
AWS Textract · OCR pipeline · Document AI
Sia Partners · Data platform
SAP → Snowflake Data Migration
ProblemBusiness-critical data
had to move from SAP to a modern warehouse without silently
corrupting anything on the way.
BuiltDelivered an end-to-end
SAP → Matillion → Snowflake migration, with validation and error
handling built into the pipeline.
ImpactData quality and pipeline
reliability protected through the cutover.
Snowflake · Matillion · Data quality
