Work.

Investor Classification System (ICS)

Problem

Sales and Account Management at an M&A marketplace had no shared, objective way to prioritize investors. Attention drifted to the loudest names, not the ones most likely to close.

Solution

A weighted scoring framework that aggregates five dimensions, Investor Type, Investment Ticket, Thesis Completion, M&A Experience, and Acquisition Time Expectation, into a continuous score in [0, 1], then discretizes it into five target levels: No Target, Low, Mid, High, Whale.

Features

• Global weights: Investor Type 0.30, Ticket 0.275, Thesis 0.20, Experience 0.05, Expectation 0.175, all explainable to Sales.
• Sub-weights per category (Search Fund, PE, Family Office, Industrial, Individual) grounded in observed close rates.
• Thesis Completion aggregates three binary signals: saved search, thesis fields complete, at least one favorite.
• Continuous score → 5 discrete tiers used directly in HubSpot for prioritization and routing.
• Reached v2.7, versioned, documented, and validated against real deal outcomes.

Churn Forecasting for Sellers (AlgoC v4)

Problem

Sellers were leaving mid-process, often past qualification, sometimes mid-LOI, with no way for Account Management to see it coming. Retention was reactive.

Solution

A rule-based and weighted seller-churn model producing P(churn) in [0, 1]. Deterministic overrides fire first (cancellation, disputes, last resource, advanced deal stage); if none apply, five behavioral factors are weighted into a probability.

Features

• Strict precedence hierarchy: cancellation → new-user module → high-signal overrides → weighted score → external-interest adjustment.
• Overrides encode operational truth: Last Resource = 0.9, Pre-LOI/LOI/DD = 0, sentiment ANXIOUS = 0.7, VERY_ANXIOUS = 0.9.
• Weighted factors: contacts activity, meetings, disputes, sentiment, and unresponsiveness.
• Drove seller churn from 28% → 5%, now stable around 10%.

Churn Algorithm, Investor Side

Problem

Investor churn was only understood after cancellation. Reasons were narrated, not measured. Retention interventions were based on intuition.

Solution

A five-module predictive architecture that treats investor churn as a lifecycle risk, not an event: Sentiment, Response Friction, Dealflow Fit / Reciprocity, Resilience, and Search & Thesis Complexity.

Features

• Each module captures a distinct predictive dimension, emotional signals, marketplace friction, thesis-supply mismatch, engagement decay.
• Designed to surface leading indicators early enough to intervene, manage expectations, or re-match the investor.
• First formal inventory of variables for the Data team, the contract between operational knowledge and future ML.

Sentiment Global (Companies)

Problem

Sentiment across the seller lifecycle was tracked inconsistently, a Post-Onboarding comment weighed the same as a First-Meeting red flag. No shared scale, no time decay, no playbook triggers.

Solution

A dynamic sentiment score in [0, 1] with weights that redistribute by lifecycle phase (not all categories exist at all times), robust time-aware treatment of the repeatable First Meeting, and playbook activation by bucket.

Features

• Four categories: Sales, Post-Onboarding, Post-Valuation, First Meeting (repeatable).
• EMA over First Meeting scores for recency, plus history penalty and inactivity weight decay.
• Safety constraint that protects Sales weight so it can never collapse to zero.
• Buckets → concrete playbooks so operators know exactly what to do at each risk level.

Investor Onboarding, Standardized Script

Problem

Investor onboarding calls varied wildly by operator. Portfolio retention hovered anywhere between 40% and 90% depending on who ran the meeting.

Solution

A structured 30/60-minute onboarding script covering mandatory pre-call research, thesis capture, platform walkthrough, and next steps, with a checklist for HubSpot, LinkedIn, prior interactions and internal metrics.

Features

• Pre-call block: prior calls, emails, objections, urgency signals, HubSpot state.
• External block: LinkedIn (personal + firm), corporate site, historic investments, public narrative.
• Internal metrics block: profile completeness, saved searches, favorites, hides, connections, Investor Type, Target Level.
• Each of 3 onboarded team members reached ~89% portfolio retention using the script.

Amaru Invest

Problem

Value investing is the most proven wealth-building methodology in the history of financial markets, and one of the most time-intensive. Applying multiple frameworks rigorously to a single stock can take 4-6 hours. Individual investors cut corners, apply one framework, or default to consensus opinion that is already priced in. Emotional decisions, loss aversion, FOMO, anchoring, do the rest.

Solution

A web platform that scores any stock against the criteria of 9 legendary value investors, Graham, Buffett, Munger, Lynch, Greenblatt, Fisher, Templeton, Marks, Schloss, in seconds. The output is the Graham Score: a transparent, multi-dimensional 0-100 score with each pass/fail visible, plus a paper portfolio to simulate composition and an embeddable widget for third parties.

Features

• Graham Score Engine, 9 frameworks applied simultaneously, every criterion pass/fail exposed and explained.
• Paper Portfolio, simulated positions with aggregate portfolio Graham Score and per-investor approval distribution.
• Embeddable widget (B2B2C), one-line integration for newsletters, fintechs, and content creators.
• Roadmap: earnings-call AI summaries, contrarian screener, score trend over time, Ara voice coaching agent, sector discovery, public API, mobile.
• TAM $2.5B · SAM $380M · 12% CAGR to 2033.

amaru-invest.com →

Hermes Fitness

Problem

Most fitness apps optimize for content, streaks, or aesthetics, not for the daily discipline that actually produces a transformation. Personalized plans are generic, tracking is a chore, and accountability disappears the moment life gets busy.

Solution

A training platform for athletes who want to train the way the ancients trained, with structure, progression, and community accountability. Personalized daily, weekly and monthly plans that adapt to logged performance, plus a global ranking system that keeps consistency visible.

Features

• Personalized plans, daily, weekly, monthly, tailored to declared goals and adjusted from what you actually lift.
• Rep-by-rep tracking, sets, reps, weight, with the plan re-adapting as you progress.
• Global ranking on streaks and consistency, not vanity metrics.
• Community layer, meet athletes, share the journey, push each other forward.
• Manifesto: no excuses, no fears, no limits, no pretense, with respect.

hermes-fitness.com →