Akhil Cherian Verghese

For a high-demand aesthetic practice, Krazimo built a unified operations layer (leads, scheduling, comms, call transcriptions and summaries, consents, and reporting) plus an AI concierge for Instagram and live chat—reducing lead leakage and turning high-volume inbound interest into structured intake and faster follow-ups without sacrificing a premium patient experience. The system is designed to capture intent the moment it arrives, route it correctly, and keep the patient journey consistent across channels—so the clinic isn’t dependent on staff being instantly available. With better automation around follow-ups and status tracking, the practice can move faster while still keeping conversations personal and on-brand.

PitchMee turns sales enablement into measurable practice by combining AI voice simulations (“AI Battles”), peer roleplays, and real meeting analysis—so reps can get realistic reps in, and managers get scalable coaching without listening to every call. The experience is built to tighten the feedback loop: reps practice objection handling, discovery, and closing in a safe environment, then get structured coaching based on performance patterns. The result is faster ramp time, more consistent messaging, and a training system that scales across teams and geographies.

CaseLogic is a “trust-by-architecture” legal AI workspace built for auditability: persistent case workspaces, citation-enforced retrieval, specialist multi-agent reasoning (lawyer, paralegal, co-counsel, and judge modes), and a verification layer designed to prevent confident hallucinations in high-stakes legal workflows. The system prioritizes traceability—so outputs can be reviewed, cited, and defended—rather than just sounding persuasive. That makes it practical for legal research and drafting, where accuracy and source grounding are non-negotiable, and where teams need an assistant that behaves more like a disciplined junior associate than a creative writer.

GraphAI makes on-chain analytics accessible by letting users ask natural-language questions and get safe, real-time insights—removing the need for custom indexers, ABI/RPC expertise, or rigid dashboards, especially as L2 activity and data volume explode. It bridges the gap between raw blockchain data and decision-making by translating intent into queries and returning answers that are actually usable by non-specialists. The payoff is speed and accessibility: teams can investigate wallets, protocols, and trends without waiting on a data engineer to build a bespoke report.

Blink Concierge deploys production-grade voice agents that don’t just “talk,” they complete workflows (bookings, lookups, updates, and routing) on real phone numbers—fully automating routine calls and driving human involvement as low as 2–3% for simple, repeatable businesses. It’s built like a real telephony product: reliability, call logging, routing rules, and integrations matter as much as the model. The outcome is fewer missed calls, faster after-hours responses, and a front desk that scales without hiring—while still escalating edge cases to humans when needed.

Krazimo prototyped an AI platform that compresses the slowest part of decarbonization consulting—turning scattered public and client data into a structured emissions and asset view, generating evidence-linked recommendations, dashboards, and a client-ready report so consulting scale doesn’t require proportional headcount. Instead of manually assembling research, the platform accelerates the pathway from “information gathering” to “plan creation,” with traceable sources behind key claims. This enables consultants and sustainability teams to iterate faster, explore scenarios, and produce clearer deliverables—while keeping the work grounded in defensible data.

Krazimo helped JSTFYD equip restoration teams to win insurer and TPA disputes with fast, auditable, standards-backed justifications—dropping response time from days to hours, improving settlements by around 6%, and saving around 45 hours per week for one restoration company (≈$800,000 projected annual impact). The platform centralizes the evidence trail behind every line item, making it easier to defend scope and pricing with documentation that insurers recognize. Practically, this means fewer back-and-forth cycles, less manual writing, and a cleaner process for field teams who just want to close claims quickly and move to the next job.

I was responsible for designing and building the first version of the ClearTax Android app, which received a 4.2 rating on the Play Store (with over 100,000 installs). There was one developer working under me during this project.

https://goo.gl/STfA4E

While employed at IBM, I was part of the team responsible for the development of the MaaS360 for Android app. I cannot overstate the complexity of this application. I believe the unique value I offer as an Android developer stems from the training I received at IBM.

MaaS360 is an MDM (mobile device management) provider, which means that the app controls every aspect of the device. Even when minimized, it essentially acts as a second operating system, restricting certain actions and protecting your data.

It also had several secondary applications (Docs, Email, and more). Since the app was far too large to be installed together, each component was designed separately, and I was partly responsible for ensuring smooth communication between the various sub-applications.

This was the "documents" side of the application. It involved linking the SDKs of Google Drive, Box, Dropbox, etc., so that our users could access their company files in our secure container. As I alluded to in the MaaS360 project section, this required knowledge of industry best practices in efficiency, network requests, multithreading, and security.

Although I wasn't involved in the initial design of this product, I was asked to help improve its usability and efficiency. It was one of the most used components of our suite of services, and it was receiving poor reviews on the app store, mainly because it drained our users' batteries. After analyzing the code, removing unnecessary network calls, and adding a series of features to improve usability, including multidirectional swipe (see my Android Swipe Layout project), I was able to reduce battery drain by roughly 60% during email synchronization and make the app more user-friendly.

I worked on the simulation of a patient's gait using multi-camera reconstruction and data-smoothing algorithms, and generated the angle and torque graphs for each joint. Movement was measured using LEDs attached to strategic points along the patient's legs.

I used pressure sensors, an Arduino, MATLAB, a small C++ script, ROSbridge, and HTML5 to develop a cost-effective foot pressure-mapping system that provides reasonable estimates of pressure distribution with minimal sensors.

A fairly significant contribution to one of the most popular Android swipe libraries on the internet. Essentially, I added a multidirectional swipe. The feature had been requested numerous times, but the creator thought the change would require considerable effort. After looking at the code, I realized there was a simple way to do it.

You can follow the conversation here - https://github.com/daimajia/AndroidSwipeLayout/issues/40 (my handle is harshevilgeek)

A Python script to open Android ZIP files and look for pre-programmed statements and uncaught exceptions, and to perform an ANR analysis. One of the coolest things about this script is the ability to identify cyclic dependencies and detect deadlocks, which anyone who's analyzed an ANR log can tell you, is not at all easy to do as a human!