neesh Inc.

Work

A senior team that has shipped together for years, one accountable person on every engagement, available now. What we have built, and what an AI deployment looks like.

TRACK RECORD

What Our Team Has Shipped

Work our people delivered at the firms named, shown to say what we’ve done, not as clients of neesh.

  1. FUND SERVICES

    Six legacy applications rebuilt. Releases cut from hours to minutes.

    Six of a London fund-services group’s fourteen applications, rebuilt from monoliths into services on a modern stack; one shared design system across all fourteen; a deployment pipeline that took releases from hours to minutes. A nine-person team, built and led for the work, demoed to the C-suite after every increment.

    Delivered by Shan Peiris with Lakshitha, Saranga and Aravinda at MJ Hudson (now part of Apex Group), 2021–2023

  2. AUTOMOTIVE FINANCE

    Automotive-finance systems moved to the cloud, with security checked on every change.

    Enterprise applications for an automotive-finance platform, migrated from on-premises servers to AWS. Security scanning, automated browser testing and AI-assisted code review run in the pipeline, so problems surface before a release rather than after it, and production monitoring watches what ships.

    Delivered by Kapila at Dealertrack Canada (Cox Automotive), since 2019

  3. GOVERNMENT

    A government statutory board and the Accountant-General’s department, one payment-status framework between them.

    A payment-status framework connecting Singapore’s Building and Construction Authority, a statutory board, with the Accountant-General’s department, so a payment’s status is tracked between the two rather than chased by email.

    Delivered by Shan Peiris for Singapore’s Building and Construction Authority

  4. FINANCIAL REPORTING

    The reports people waited on, turned into live dashboards.

    Real-time financial reporting dashboards for a fund-services group, and the services behind them. For an ESG reporting product, a legacy system rebuilt on a modern front-end stack.

    Delivered by Saranga at MJ Hudson and Holtara (Apex Group)

  5. TRADING & AUTOMATION

    Trading platforms engineered, and repetitive work handed to software robots.

    Trading and brokerage platforms engineered in Java. For an enterprise client, a robotic process automation program that took repetitive work off people’s desks.

    Delivered by Lakshitha at B2BROKER and Virtusa

  6. PRODUCT DESIGN

    Design systems that made enterprise products feel like one.

    Design systems and UI kits that unified enterprise product suites, built in Figma for a clean handoff to engineering. Enterprise UX audits, and front ends modernized from the design up.

    Delivered by Aravinda at Holtara (Apex Group), OpusXenta and 99X

SCENARIOS

What It Looks Like Deployed

Your senior staff answer the same 30 questions. Every single week.

CHALLENGE_VECTOR 165 employees. 8 years of institutional knowledge scattered across Confluence, Drive, and S3. Senior partners fielding 30+ interruptions weekly.
SOLUTION_APPROACH Permission-aware RAG connected to all three document sources. Answers in Slack, every response cited and traceable, with a design target of under 10 seconds.
OUTCOME_INDEX 73% fewer knowledge interruptions. New hire time-to-productivity: 8 weeks to 3. 14 undocumented processes surfaced by gap detection. DESIGN TARGETS · NOT CUSTOMER RESULTS
BASE_PRODUCT AskBase · available now
CAPABILITY_SET
PERMISSION-AWARE RAGSLACKGAP DETECTION
PythonFastAPIPostgreSQL + pgvectorClaude HaikuClaude SonnetConfluence APIGoogle Drive APISlack BoltAWS ECS Fargate

Your property management inbox handles 68% of emails. Automatically.

CHALLENGE_VECTOR 240 units. 60+ emails per day. Two property managers spending 90 minutes every morning just triaging, before a single response was written.
SOLUTION_APPROACH FlowBase deployment that classifies every email on arrival, drafts replies to routine requests for a person to approve, queues borderline cases, and escalates urgent or legal triggers.
OUTCOME_INDEX 68% of emails handled without human input. Average response time: 3.5 hours to 4 minutes. 11 hours saved per week. DESIGN TARGETS · NOT CUSTOMER RESULTS
BASE_PRODUCT FlowBase · in build
CAPABILITY_SET
LLM TRIAGECONFIDENCE ROUTINGEMAILLLM-AS-JUDGEREVIEW QUEUE
Microsoft GraphFastAPIPostgreSQLClaude HaikuClaude SonnetGmail APIAWS ECS Fargate

Your competitor changed their pricing. You found out three days later.

CHALLENGE_VECTOR Four direct competitors. A buying manager with a bookmark folder and a spreadsheet, spending 45 minutes every morning checking sites manually, and missing the moves that happened overnight.
SOLUTION_APPROACH PriceWatch deployed across 8 competitor pages. Local ML filters noise at zero cost, and the LLM narrates only the changes that clear the threshold. A high change reaches Slack the moment it is found; a weekly briefing follows every Monday.
OUTCOME_INDEX 4 days pre-warning on Black Friday pricing. First competitive response in under 2 hours. AI cost for the week that caught it: $0.04. DESIGN TARGETS · NOT CUSTOMER RESULTS
BASE_PRODUCT WatchBase · in build
CAPABILITY_SET
WEB MONITORINGML SIGNIFICANCE SCORINGSEMANTIC CACHELLM NARRATIONSLACK BRIEFINGS
PythonFastAPIPostgreSQL + pgvectorClaude HaikuClaude SonnetHuggingFace TransformersPlaywrightAWS ECS Fargate

Your best proposals take 12 hours to write. This one took 3 minutes.

CHALLENGE_VECTOR Every proposal starts from a blank document. Quality varies by deadline pressure, not by what the firm knows. The institutional memory that wins pitches does not accumulate.
SOLUTION_APPROACH Five specialist agents running in parallel, a review pass for cross-section consistency, and a quality gate that scores every document before delivery.
OUTCOME_INDEX Complete draft in under 3 minutes. Every document scored by the judge before delivery, and revised when the judge does not pass it (its pass mark is 7.0 / 10). LLM cost under $0.40 per document. DESIGN TARGETS · NOT CUSTOMER RESULTS
BASE_PRODUCT DocBase · in build
CAPABILITY_SET
MULTI-AGENT GENERATIONQUALITY GATESSTRUCTURED OUTPUTTEMPLATESPDF / DOCX
PythonFastAPIPostgreSQL + pgvectorClaude HaikuClaude SonnetClaude OpusNext.jsAWS ECS Fargate

Have a problem worth solving?
Let's talk.

Tell us what's eating your team's time. We'll show you what a machine can do instead.

Book Free Assessment